Neural ranking models (NRMs) have demonstrated effective performance in several information retrieval (IR) tasks. However, training NRMs often requires large-scale training data, which is difficult and expensive to obtain. To address this issue, one can train NRMs via weak supervision, where a large dataset is automatically generated using an existing ranking model (called the weak labeler) for training NRMs. Weakly supervised NRMs can generalize from the observed data and significantly outperform the weak labeler. This paper generalizes this idea through an iterative re-labeling process, demonstrating that weakly supervised models can iteratively play the role of weak labeler and significantly improve ranking performance without using manually labeled data. The proposed Generalized Weak Supervision (GWS) solution is generic and orthogonal to the ranking model architecture. This paper offers four implementations of GWS: self-labeling, cross-labeling, joint cross- and self-labeling, and greedy multi-labeling. GWS also benefits from a query importance weighting mechanism based on query performance prediction methods to reduce noise in the generated training data. We further draw a theoretical connection between self-labeling and Expectation-Maximization. Our experiments on four retrieval benchmarks suggest that our implementations of GWS lead to substantial improvements compared to weak supervision if the weak labeler is sufficiently reliable.

Learning multiple intent representations for queries has potential applications in facet generation, document ranking, search result diversification, and search explanation. The state-of-the-art model for this task assumes that there is a sequence of intent representations. In this paper, we argue that the model should not be penalized as long as it generates an accurate and complete set of intent representations. Based on this intuition, we propose a stochastic permutation invariant approach for optimizing such networks. We extrinsically evaluate the proposed approach on a facet generation task and demonstrate significant improvements compared to competitive baselines. Our analysis shows that the proposed permutation invariant approach has the highest impact on queries with more potential intents.

Non-factoid question answering (NFQA) is a challenging and under-researched task that requires constructing long-form answers, such as explanations or opinions, to open-ended non-factoid questions - NFQs. There is still little understanding of the categories of NFQs that people tend to ask, what form of answers they expect to see in return, and what the key research challenges of each category are.

Abstract This research analyzes human‐generated clarification questions to provide insights into how they are used to disambiguate and provide a better understanding of information needs. A set of clarification questions is extracted from posts on the Stack Exchange platform. Novel taxonomy is defined for the annotation of the questions and their responses. We investigate the clarification questions in terms of whether they add any information to the post (the initial question posted by the asker) and the accepted answer, which is the answer chosen by the asker. After identifying, which clarification questions are more useful, we investigated the characteristics of these questions in terms of their types and patterns. Non‐useful clarification questions are identified, and their patterns are compared with useful clarifications. Our analysis indicates that the most useful clarification questions have similar patterns, regardless of topic. This research contributes to an understanding of clarification in conversations and can provide insight for clarification dialogues in conversational search scenarios and for the possible system generation of clarification requests in information‐seeking conversations.

Taking a user-centric approach, we study the features that render an answer to a non-factoid question useful in the eyes of the person who asked that question. An editorial study, where participants assess the usefulness of the answers they received in response to their questions, as well as 12 different aspects associated with the answers, indicates considerable correlation between certain aspects such as relevance, correctness, and completeness with the user-perceived usefulness of answers. Moreover, we investigate the effectiveness of some commonly used answer quality measures, such as ROGUE, BLEU, METEOR, and BERTScore, demonstrating that these measures are limited in their ability to capture the aspects of usefulness and have room for improvement. The question answering dataset created in our work was made publicly available.

We present a new, multi-faceted taxonomy to classify questions asked in web search engines based on the question intent, types of entities mentioned, types of question words, and granularity of the expected answer. Built based on the inspection of 1,000 real-life questions issued to a web search engine, the taxonomy reflects the recent search behavior of users and enables deep understanding of user intents, goals, and expected answers. This taxonomy is more fine-grained than previous query taxonomies, and is designed with the ultimate goal of reducing the inherent ambiguity in determining the intent of questions. In addition, we describe the formal procedure for conducting an editorial study of the taxonomy including its evaluation. The adopted procedure aims to increase assessor agreement without incurring too much overhead. Our results demonstrate that, despite being more fine-grained, the proposed intent categories result in higher agreement between assessors compared to an existing, commonly used taxonomy.

Existing commercial search engines often struggle to represent different perspectives of a search query. Argument retrieval systems address this limitation of search engines and provide both positive (PRO) and negative (CON) perspectives about a user's information need on a controversial topic (e.g., climate change). The effectiveness of such argument retrieval systems is typically evaluated based on topical relevance and argument quality, without taking into account the often differing number of documents shown for the argument stances (PRO or CON). Therefore, systems may retrieve relevant passages, but with a biased exposure of arguments. In this work, we analyze a range of non-stochastic fairness-aware ranking and diversity metrics to evaluate the extent to which argument stances are fairly exposed in argument retrieval systems.

Representation learning has always played an important role in information retrieval (IR) systems. Most retrieval models, including recent neural approaches, use representations to calculate similarities between queries and documents to find relevant information from a corpus. Recent models use large-scale pre-trained language models for query representation. The typical use of these models, however, has a major limitation in that they generate only a single representation for a query, which may have multiple intents or facets. The focus of this paper is to address this limitation by considering neural models that support multiple intent representations for each query. Specifically, we propose the NMIR (Neural Multiple Intent Representations) model that can generate semantically different query intents and their appropriate representations. We evaluate our model on query facet generation using a large-scale dataset of real user queries sampled from the Bing search logs. We also provide an extrinsic evaluation of the proposed model using a clarifying question selection task. The results show that NMIR significantly outperforms competitive baselines.

Product search has been a crucial entry point to serve people shopping online. Most existing personalized product models follow the paradigm of representing and matching user intents and items in the semantic space, where finer-grained matching is totally discarded and the ranking of an item cannot be explained further than just user/item level similarity. In addition, while some models in existing studies have created dynamic user representations based on search context, their representations for items are static across all search sessions. This makes every piece of information about the item always equally important in representing the item during matching with various user intents. Aware of the above limitations, we propose a review-based transformer model (RTM) for personalized product search, which encodes the sequence of query, user reviews, and item reviews with a transformer architecture. RTM conducts review-level matching between the user and item, where each review has a dynamic effect according to the context in the sequence. This makes it possible to identify useful reviews to explain the scoring. Experimental results show that RTM significantly outperforms state-of-the-art personalized product search baselines.

In this work, we address multi-modal information needs that contain text questions and images by focusing on passage retrieval for outside-knowledge visual question answering. This task requires access to outside knowledge, which in our case we define to be a large unstructured passage collection. We first conduct sparse retrieval with BM25 and study expanding the question with object names and image captions. We verify that visual clues play an important role and captions tend to be more informative than object names in sparse retrieval. We then construct a dual-encoder dense retriever, with the query encoder being LXMERT, a multi-modal pre-trained transformer. We further show that dense retrieval significantly outperforms sparse retrieval that uses object expansion. Moreover, dense retrieval matches the performance of sparse retrieval that leverages human-generated captions.

Users install many apps on their smartphones, raising issues related to information overload for users and resource management for devices. Moreover, the recent increase in the use of personal assistants has made mobile devices even more pervasive in users’ lives. This article addresses two research problems that are vital for developing effective personal mobile assistants: target apps selection and recommendation . The former is the key component of a unified mobile search system: a system that addresses the users’ information needs for all the apps installed on their devices with a unified mode of access. The latter, instead, predicts the next apps that the users would want to launch. Here we focus on context-aware models to leverage the rich contextual information available to mobile devices. We design an in situ study to collect thousands of mobile queries enriched with mobile sensor data (now publicly available for research purposes). With the aid of this dataset, we study the user behavior in the context of these tasks and propose a family of context-aware neural models that take into account the sequential, temporal, and personal behavior of users. We study several state-of-the-art models and show that the proposed models significantly outperform the baselines.

To study the impact of providing direct answers in search results on user behavior, we conducted a controlled user study to analyze factors including reading time, eye-tracked attention, and the influence of the quality of answer module content. We also studied a more advanced answer interface, where multiple answers are shown on the search engine results page (SERP). Our results show that users focus more extensively than normal on the top items in the result list when answers are provided. The existence of the answer module helps to improve user engagement on SERPs, reduces user effort, and promotes user satisfaction during the search process. Furthermore, we investigate how the question type -- factoid or non-factoid -- affects user interaction patterns. This work provides insight into the design of SERPs that includes direct answers to queries, including when answers should be shown.

We investigated how users evaluate passage-length answers for non-factoid questions. We conduct a study where answers were presented to users, sometimes shown with automatic word highlighting. Users were tasked with evaluating answer quality, correctness, completeness, and conciseness. Words in the answer were also annotated, both explicitly through user mark up and implicitly through user gaze data obtained from eye-tracking. Our results show that the correctness of an answer strongly depends on its completeness, conciseness is less important.

Product search is an important way for people to browse and purchase items on E-commerce platforms. While customers tend to make choices based on their personal tastes and preferences, analysis of commercial product search logs has shown that personalization does not always improve product search quality. Most existing product search techniques, however, conduct undifferentiated personalization across search sessions. They either use a fixed coefficient to control the influence of personalization or let personalization take effect all the time with an attention mechanism. The only notable exception is the recently proposed zero-attention model (ZAM) that can adaptively adjust the effect of personalization by allowing the query to attend to a zero vector. Nonetheless, in ZAM, personalization can act at most as equally important as the query and the representations of items are static across the collection regardless of the items co-occurring in the user's historical purchases. Aware of these limitations, we propose a transformer-based embedding model (TEM) for personalized product search, which could dynamically control the influence of personalization by encoding the sequence of query and user's purchase history with a transformer architecture. Personalization could have a dominant impact when necessary and interactions between items can be taken into consideration when computing attention weights. Experimental results show that TEM outperforms state-of-the-art personalization product retrieval models significantly.

Asking clarifying questions in response to ambiguous or faceted queries has been recognized as a useful technique for various information retrieval systems, especially conversational search systems with limited bandwidth interfaces. Analyzing and generating clarifying questions have been studied recently but the accurate utilization of user responses to clarifying questions has been relatively less explored. In this paper, we enrich the representations learned by Transformer networks using a novel attention mechanism from external information sources that weights each term in the conversation. We evaluate this Guided Transformer model in a conversational search scenario that includes clarifying questions. In our experiments, we use two separate external sources, including the top retrieved documents and a set of different possible clarifying questions for the query. We implement the proposed representation learning model for two downstream tasks in conversational search; document retrieval and next clarifying question selection. Our experiments use a public dataset for search clarification and demonstrate significant improvements compared to competitive baselines.

Recipe retrieval is a representative and useful application of cross-modal information retrieval. Recent studies have proposed frameworks for retrieving images of cuisines given textual ingredient lists and instructions. However, the textual form of ingredients easily causes information loss or inaccurate description, especially for novices of cookery who are often the main users of recipe retrieval systems. In this paper, we revisit the task of recipe retrieval by taking images of ingredients as input queries, and retrieving cuisine images by incorporating visual information of ingredients through a deep convolutional neural network. We build an image-to-image recipe retrieval system to validate the effect of ingredient image queries. We further combine the proposed solution with a state-of-the-art cross-modal recipe retrieval model to improve the overall performance of the recipe retrieval task.

Conversational search is one of the ultimate goals of information retrieval. Recent research approaches conversational search by simplified settings of response ranking and conversational question answering, where an answer is either selected from a given candidate set or extracted from a given passage. These simplifications neglect the fundamental role of retrieval in conversational search. To address this limitation, we introduce an open-retrieval conversational question answering (ORConvQA) setting, where we learn to retrieve evidence from a large collection before extracting answers, as a further step towards building functional conversational search systems. We create a dataset, OR-QuAC, to facilitate research on ORConvQA. We build an end-to-end system for ORConvQA, featuring a retriever, a reranker, and a reader that are all based on Transformers. Our extensive experiments on OR-QuAC demonstrate that a learnable retriever is crucial for ORConvQA. We further show that our system can make a substantial improvement when we enable history modeling in all system components. Moreover, we show that the reranker component contributes to the model performance by providing a regularization effect. Finally, further in-depth analyses are performed to provide new insights into ORConvQA.

In this work, we focus on the contextual document ranking task, which deals with the challenge of user interaction modeling for conversational search. Given a history of user feedback behaviors, such as issuing a query, clicking a document, and skipping a document, we propose to introduce behavior awareness to a neural ranker, resulting in a Hierarchical Behavior Aware Transformers (HBA-Transformers) model. The hierarchy is composed of an intra-behavior attention layer and an inter-behavior attention layer to let the system effectively distinguish and model different user behaviors. Our extensive experiments on the AOL session dataset demonstrate that the hierarchical behavior aware architecture is more powerful than a simple combination of history behaviors. Besides, we analyze the conversational property of queries. We show that coherent sessions tend to be more conversational and thus are more demanding in terms of considering history user behaviors.

Existing learning to rank models for information retrieval are trained based on explicit or implicit query-document relevance information. In this paper, we study the task of learning a retrieval model based on user-item interactions. Our model has potential applications to the systems with rich user-item interaction data, such as browsing and recommendation, in which having an accurate search engine is desired. This includes media streaming services and e-commerce websites among others. Inspired by the neural approaches to collaborative filtering and the language modeling approaches to information retrieval, our model is jointly optimized to predict user-item interactions and reconstruct the item textual descriptions. In more details, our model learns user and item representations such that they can accurately predict future user-item interactions, while generating an effective unigram language model for each item. Our experiments on four diverse datasets in the context of movie and product search and recommendation demonstrate that our model substantially outperforms competitive retrieval baselines, in addition to providing comparable performance to state-of-the-art hybrid recommendation models.

Personal assistant systems, such as Apple Siri, Google Assistant, Amazon Alexa, and Microsoft Cortana, are becoming ever more widely used. Understanding user intent such as clarification questions, potential answers and user feedback in information-seeking conversations is critical for retrieving good responses. In this paper, we analyze user intent patterns in information-seeking conversations and propose an intent-aware neural response ranking model “IART”, which refers to “Intent-Aware Ranking with Transformers”. IART is built on top of the integration of user intent modeling and language representation learning with the Transformer architecture, which relies entirely on a self-attention mechanism instead of recurrent nets [35]. It incorporates intent-aware utterance attention to derive an importance weighting scheme of utterances in conversation context with the aim of better conversation history understanding. We conduct extensive experiments with three information-seeking conversation data sets including both standard benchmarks and commercial data. Our proposed model outperforms all baseline methods with respect to a variety of metrics. We also perform case studies and analysis of learned user intent and its impact on response ranking in information-seeking conversations to provide interpretation of results.

Product search is one of the most popular methods for customers to discover products online. Most existing studies on product search focus on developing effective retrieval models that rank items by their likelihood to be purchased. However, they ignore the problem that there is a gap between how systems and customers perceive the relevance of items. Without explanations, users may not understand why product search engines retrieve certain items for them, which consequentially leads to imperfect user experience and suboptimal system performance in practice. In this work, we tackle this problem by constructing explainable retrieval models for product search. Specifically, we propose to model the “search and purchase” behavior as a dynamic relation between users and items, and create a dynamic knowledge graph based on both the multi-relational product data and the context of the search session. Ranking is conducted based on the relationship between users and items in the latent space, and explanations are generated with logic inferences and entity soft matching on the knowledge graph. Empirical experiments show that our model, which we refer to as the Dynamic Relation Embedding Model (DREM), significantly outperforms the state-of-the-art baselines and has the ability to produce reasonable explanations for search results.

Information retrieval systems are evolving from document retrieval to answer retrieval. Web search logs provide large amounts of data about how people interact with ranked lists of documents, but very little is known about interaction with answer texts. In this paper, we use Amazon Mechanical Turk to investigate three answer presentation and interaction approaches in a non-factoid question answering setting. We find that people perceive and react to good and bad answers very differently, and can identify good answers relatively quickly. Our results provide the basis for further investigation of effective answer interaction and feedback methods.

Conversational assistants are being progressively adopted by the general population. However, they are not capable of handling complicated information-seeking tasks that involve multiple turns of information exchange. Due to the limited communication bandwidth in conversational search, it is important for conversational assistants to accurately detect and predict user intent in information-seeking conversations. In this paper, we investigate two aspects of user intent prediction in an information-seeking setting. First, we extract features based on the content, structural, and sentiment characteristics of a given utterance, and use classic machine learning methods to perform user intent prediction. We then conduct an in-depth feature importance analysis to identify key features in this prediction task. We find that structural features contribute most to the prediction performance. Given this finding, we construct neural classifiers to incorporate context information and achieve better performance without feature engineering. Our findings can provide insights into the important factors and effective methods of user intent prediction in information-seeking conversations.

Intelligent personal assistant systems that are able to have multi-turn conversations with human users are becoming increasingly popular. Most previous research has been focused on using either retrieval-based or generation-based methods to develop such systems. Retrieval-based methods have the advantage of returning fluent and informative responses with great diversity. However, the performance of the methods is limited by the size of the response repository. On the other hand, generation-based methods can produce highly coherent responses on any topics. But the generated responses are often generic and not informative due to the lack of grounding knowledge. In this paper, we propose a hybrid neural conversation model that combines the merits of both response retrieval and generation methods. Experimental results on Twitter and Foursquare data show that the proposed model outperforms both retrieval-based methods and generation-based methods (including a recently proposed knowledge-grounded neural conversation model) under both automatic evaluation metrics and human evaluation. We hope that the findings in this study provide new insights on how to integrate text retrieval and text generation models for building conversation systems.

Product search is one of the most popular methods for people to discover and purchase products on e-commerce websites. Because personal preferences often have an important influence on the purchase decision of each customer, it is intuitive that personalization should be beneficial for product search engines. While synthetic experiments from previous studies show that purchase histories are useful for identifying the individual intent of each product search session, the effect of personalization on product search in practice, however, remains mostly unknown. In this paper, we formulate the problem of personalized product search and conduct large-scale experiments with search logs sampled from a commercial e-commerce search engine. Results from our preliminary analysis show that the potential of personalization depends on query characteristics, interactions between queries, and user purchase histories. Based on these observations, we propose a Zero Attention Model for product search that automatically determines when and how to personalize a user-query pair via a novel attention mechanism. Empirical results on commercial product search logs show that the proposed model not only significantly outperforms state-of-the-art personalized product retrieval models, but also provides important information on the potential of personalization in each product search session.

Intelligent assistants change the way people interact with computers and make it possible for people to search for products through conversations when they have purchase needs. During the interactions, the system could ask questions on certain aspects of the ideal products to clarify the users' needs. For example, previous work proposed to ask users the exact characteristics of their ideal items before showing results. However, users may not have clear ideas about what an ideal item looks like, especially when they have not seen any item. So it is more feasible to facilitate the conversational search by showing example items and asking for feedback instead. In addition, when the users provide negative feedback for the presented items, it is easier to collect their detailed feedback on certain properties (aspect-value pairs) of the non-relevant items. By breaking down the item-level negative feedback to fine-grained feedback on aspect-value pairs, more information is available to help clarify users' intents. So in this paper, we propose a conversational paradigm for product search driven by non-relevant items, based on which fine-grained feedback is collected and utilized to show better results in the next iteration. We then propose an aspect-value likelihood model to incorporate both positive and negative feedback on fine-grained aspect-value pairs of the non-relevant items. Experimental results show that our model is significantly better than state-of-the-art product search baselines without using feedback and those baselines using item-level negative feedback.

In product search, users tend to browse results on multiple search result pages (SERPs) (e.g., for queries on clothing and shoes) before deciding which item to purchase. Users' clicks can be considered as implicit feedback which indicates their preferences and used to re-rank subsequent SERPs. Relevance feedback (RF) techniques are usually involved to deal with such scenarios. However, these methods are designed for document retrieval, where relevance is the most important criterion. In contrast, product search engines need to retrieve items that are not only relevant but also satisfactory in terms of customers' preferences. Personalization based on users' purchase history has been shown to be effective in product search. However, this method captures users' long-term interest, which do not always align with their short-term interest, and does not benefit customers with little or no purchase history. In this paper, we study RF techniques based on both long-term and short-term context dependencies in multi-page product search. We also propose an end-to-end context-aware embedding model which can capture both types of context. Our experimental results show that short-term context leads to much better performance compared with long-term and no context. Moreover, our proposed model is more effective than state-of-art word-based RF models.

Users are known to interact more with fresh content in certain temporally associated domains such as news search or job seeking, leading to an uneven distribution of interactions over items of different degrees of freshness. Data collected under such an "aging effect'' is usually used unconditionally on all sort of recommendation tasks, and as a result more recently published content may be over-represented during model training and evaluation. In this study, we characterize this temporal influence as a recency bias, and present an analysis in the domain of job recommendation. We show that, by correcting for recency bias using an unbiased learning to rank approach, one can improve the quality of recommendation significantly over a recent neural collaborative filtering model on RecSys Challenge 2017 data.

Conversational question answering (ConvQA) is a simplified but concrete setting of conversational search. One of its major challenges is to leverage the conversation history to understand and answer the current question. In this work, we propose a novel solution for ConvQA that involves three aspects. First, we propose a positional history answer embedding method to encode conversation history with position information using BERT in a natural way. BERT is a powerful technique for text representation. Second, we design a history attention mechanism (HAM) to conduct a "soft selection" for conversation histories. This method attends to history turns with different weights based on how helpful they are on answering the current question. Third, in addition to handling conversation history, we take advantage of multi-task learning (MTL) to do answer prediction along with another essential conversation task (dialog act prediction) using a uniform model architecture. MTL is able to learn more expressive and generic representations to improve the performance of ConvQA. We demonstrate the effectiveness of our model with extensive experimental evaluations on QuAC, a large-scale ConvQA dataset. We show that position information plays an important role in conversation history modeling. We also visualize the history attention and provide new insights into conversation history understanding.

Users often fail to formulate their complex information needs in a single query. As a consequence, they may need to scan multiple result pages or reformulate their queries, which may be a frustrating experience. Alternatively, systems can improve user satisfaction by proactively asking questions of the users to clarify their information needs. Asking clarifying questions is especially important in conversational systems since they can only return a limited number of (often only one) result(s).

There has historically been a divide between the user-oriented and system-oriented research communities in information retrieval. In my opinion, this divide is based primarily on a difference in viewpoint about the relative importance of understanding how people search for information compared to developing new retrieval models and ranking algorithms. There is strong agreement, however, that the interaction between the user and the search engine is a fundamental part of the IR process. The IR field was one of the first in computer science to recognize the importance of the user-system interaction, which led to a number of core concepts such as relevance, ranking, result presentation, feedback, evaluation, and browsing. The key message of this talk is that effective information access requires interaction between the user and the system, where both play a role. Additionally, there is growing evidence that even more effective information access can be achieved by a system that actively supports interaction, particularly in the limited-bandwidth environments provided by mobile devices and voice-based assistants. In this talk, I will first give an overview of past IR research on user-system interaction. In much of this research, the system provides passive support for the retrieval process and much of the burden for effective retrieval stays with the user. There has been some research, however, that has attempted to actively support the interaction by designing expert intermediary systems. After this review, I will focus on two current areas of research where active support for interaction is crucial. These are question answering and conversational search. These areas have recently become popular in the NLP community but they have deep roots in IR. I will describe the specific lines of research we have followed at the Center for Intelligent Information Retrieval and RMIT, including interactive answer passage retrieval, studies of information-seeking dialogues, and neural models for selecting responses and answers. Although there are many aspects to this research, I will highlight the parts where interaction is important, how we have attempted to evaluate the research, and where significant progress needs to be made.

Conversational search is an emerging topic in the information retrieval community. One of the major challenges to multi-turn conversational search is to model the conversation history to answer the current question. Existing methods either prepend history turns to the current question or use complicated attention mechanisms to model the history. We propose a conceptually simple yet highly effective approach referred to as history answer embedding. It enables seamless integration of conversation history into a conversational question answering (ConvQA) model built on BERT (Bidirectional Encoder Representations from Transformers). We first explain our view that ConvQA is a simplified but concrete setting of conversational search, and then we provide a general framework to solve ConvQA. We further demonstrate the effectiveness of our approach under this framework. Finally, we analyze the impact of different numbers of history turns under different settings to provide new insights into conversation history modeling in ConvQA.

Deep text matching approaches have been widely studied for many applications including question answering and information retrieval systems. To deal with a domain that has insufficient labeled data, these approaches can be used in a Transfer Learning (TL) setting to leverage labeled data from a resource-rich source domain. To achieve better performance, source domain data selection is essential in this process to prevent the "negative transfer" problem. However, the emerging deep transfer models do not fit well with most existing data selection methods, because the data selection policy and the transfer learning model are not jointly trained, leading to sub-optimal training efficiency. In this paper, we propose a novel reinforced data selector to select high-quality source domain data to help the TL model. Specifically, the data selector "acts" on the source domain data to find a subset for optimization of the TL model, and the performance of the TL model can provide "rewards" in turn to update the selector. We build the reinforced data selector based on the actor-critic framework and integrate it to a DNN based transfer learning model, resulting in a Reinforced Transfer Learning (RTL) method. We perform a thorough experimental evaluation on two major tasks for text matching, namely, paraphrase identification and natural language inference. Experimental results show the proposed RTL can significantly improve the performance of the TL model. We further investigate different settings of states, rewards, and policy optimization methods to examine the robustness of our method. Last, we conduct a case study on the selected data and find our method is able to select source domain data whose Wasserstein distance is close to the target domain data. This is reasonable and intuitive as such source domain data can provide more transferability power to the model.

Implicit feedback (e.g., user clicks) is an important source of data for modern search engines. While heavily biased [8, 9, 11, 27], it is cheap to collect and particularly useful for user-centric retrieval applications such as search ranking. To develop an unbiased learning-to-rank system with biased feedback, previous studies have focused on constructing probabilistic graphical models (e.g., click models) with user behavior hypothesis to extract and train ranking systems with unbiased relevance signals. Recently, a novel counterfactual learning framework that estimates and adopts examination propensity for unbiased learning to rank has attracted much attention. Despite its popularity, there is no systematic comparison of the unbiased learning-to-rank frameworks based on counterfactual learning and graphical models. In this tutorial, we aim to provide an overview of the fundamental mechanism for unbiased learning to rank. We will describe the theory behind existing frameworks, and give detailed instructions on how to conduct unbiased learning to rank in practice.

With the recent growth in the use of conversational systems and intelligent assistants such as Google Assistant and Microsoft Cortana, mobile devices are becoming even more pervasive in our lives. As a consequence, users are getting engaged with mobile apps and frequently search for an information need using different apps. Recent work has stated the need for a unified mobile search system that would act as meta search on users' mobile devices: it would identify the target apps for the user's query, submit the query to the apps, and present the results to the user. Moreover, mobile devices provide rich contextual information about users and their whereabouts. In this paper, we introduce the task of context-aware target apps selection as part of a unified mobile search framework. To this aim, we designed an in situ study to collect thousands of mobile queries enriched with mobile sensor data from 255 users during a three month period. With the aid of this dataset, we were able to study user behavior as they performed cross-app search. We finally study the performance of state-of-the-art retrieval models for this task and propose a simple yet effective neural model that significantly outperforms the baselines. Our neural approach is based on learning high-dimensional representations for mobile apps and contextual information. Furthermore, we show that incorporating context improves the performance by 20% in terms of [email protected], enabling the model to perform better for 57% of users. Our data is publicly available for research purposes.

The availability of massive data and computing power allowing for effective data driven neural approaches is having a major impact on machine learning and information retrieval research, but these models have a basic problem with efficiency. Current neural ranking models are implemented as multistage rankers: for efficiency reasons, the neural model only re-ranks the top ranked documents retrieved by a first-stage efficient ranker in response to a given query. Neural ranking models learn dense representations causing essentially every query term to match every document term, making it highly inefficient or intractable to rank the whole collection. The reliance on a first stage ranker creates a dual problem: First, the interaction and combination effects are not well understood. Second, the first stage ranker serves as a "gate-keeper" or filter, effectively blocking the potential of neural models to uncover new relevant documents. In this work, we propose a standalone neural ranking model (SNRM) by introducing a sparsity property to learn a latent sparse representation for each query and document. This representation captures the semantic relationship between the query and documents, but is also sparse enough to enable constructing an inverted index for the whole collection. We parameterize the sparsity of the model to yield a retrieval model as efficient as conventional term based models. Our model gains in efficiency without loss of effectiveness: it not only outperforms the existing term matching baselines, but also performs similarly to the recent re-ranking based neural models with dense representations. Our model can also take advantage of pseudo-relevance feedback for further improvements. More generally, our results demonstrate the importance of sparsity in neural IR models and show that dense representations can be pruned effectively, giving new insights about essential semantic features and their distributions.

Conversational search and recommendation based on user-system dialogs exhibit major differences from conventional search and recommendation tasks in that 1) the user and system can interact for multiple semantically coherent rounds on a task through natural language dialog, and 2) it becomes possible for the system to understand the user needs or to help users clarify their needs by asking appropriate questions from the users directly. We believe the ability to ask questions so as to actively clarify the user needs is one of the most important advantages of conversational search and recommendation. In this paper, we propose and evaluate a unified conversational search/recommendation framework, in an attempt to make the research problem doable under a standard formalization. Specifically, we propose a System Ask -- User Respond (SAUR) paradigm for conversational search, define the major components of the paradigm, and design a unified implementation of the framework for product search and recommendation in e-commerce. To accomplish this, we propose the Multi-Memory Network (MMN) architecture, which can be trained based on large-scale collections of user reviews in e-commerce. The system is capable of asking aspect-based questions in the right order so as to understand the user needs, while (personalized) search is conducted during the conversation, and results are provided when the system feels confident. Experiments on real-world user purchasing data verified the advantages of conversational search and recommendation against conventional search and recommendation algorithms in terms of standard evaluation measures such as NDCG.

Learning to rank has been intensively studied and widely applied in information retrieval. Typically, a global ranking function is learned from a set of labeled data, which can achieve good performance on average but may be suboptimal for individual queries by ignoring the fact that relevant documents for different queries may have different distributions in the feature space. Inspired by the idea of pseudo relevance feedback where top ranked documents, which we refer as the local ranking context, can provide important information about the query's characteristics, we propose to use the inherent feature distributions of the top results to learn a Deep Listwise Context Model that helps us fine tune the initial ranked list. Specifically, we employ a recurrent neural network to sequentially encode the top results using their feature vectors, learn a local context model and use it to re-rank the top results. There are three merits with our model: (1) Our model can capture the local ranking context based on the complex interactions between top results using a deep neural network; (2) Our model can be built upon existing learning-to-rank methods by directly using their extracted feature vectors; (3) Our model is trained with an attention-based loss function, which is more effective and efficient than many existing listwise methods. Experimental results show that the proposed model can significantly improve the state-of-the-art learning to rank methods on benchmark retrieval corpora.

Learning to rank with biased click data is a well-known challenge. A variety of methods has been explored to debias click data for learning to rank such as click models, result interleaving and, more recently, the unbiased learning-to-rank framework based on inverse propensity weighting. Despite their differences, most existing studies separate the estimation of click bias (namely the propensity model ) from the learning of ranking algorithms. To estimate click propensities, they either conduct online result randomization, which can negatively affect the user experience, or offline parameter estimation, which has special requirements for click data and is optimized for objectives (e.g. click likelihood) that are not directly related to the ranking performance of the system. In this work, we address those problems by unifying the learning of propensity models and ranking models. We find that the problem of estimating a propensity model from click data is a dual problem of unbiased learning to rank. Based on this observation, we propose a Dual Learning Algorithm (DLA) that jointly learns an unbiased ranker and an unbiased propensity model. DLA is an automatic unbiased learning-to-rank framework as it directly learns unbiased ranking models from biased click data without any preprocessing. It can adapt to the change of bias distributions and is applicable to online learning. Our empirical experiments with synthetic and real-world data show that the models trained with DLA significantly outperformed the unbiased learning-to-rank algorithms based on result randomization and the models trained with relevance signals extracted by click models.

With the recent growth of conversational systems and intelligent assistants such as Apple Siri and Google Assistant, mobile devices are becoming even more pervasive in our lives. As a consequence, users are getting engaged with the mobile apps and frequently search for an information need in their apps. However, users cannot search within their apps through their intelligent assistants. This requires a unified mobile search framework that identifies the target app(s) for the user's query, submits the query to the app(s), and presents the results to the user. In this paper, we take the first step forward towards developing unified mobile search. In more detail, we introduce and study the task of target apps selection, which has various potential real-world applications. To this aim, we analyze attributes of search queries as well as user behaviors, while searching with different mobile apps. The analyses are done based on thousands of queries that we collected through crowdsourcing. We finally study the performance of state-of-the-art retrieval models for this task and propose two simple yet effective neural models that significantly outperform the baselines. Our neural approaches are based on learning high-dimensional representations for mobile apps. Our analyses and experiments suggest specific future directions in this research area.

Learning to rank is a key component of modern information retrieval systems. Recently, regression forest models (i.e., random forests, LambdaMART and gradient boosted regression trees) have come to dominate learning to rank systems in practice, as they provide the ability to learn from large scale data while generalizing well to additional test queries. As a result, efficient implementations of these models is a concern in production systems, as evidenced by past work. We propose an alternate method for optimizing the execution of learned models: converting these expensive ensembles to a feed-forward neural network. This simple neural architecture is quite efficient to execute: we show that the resulting chain of matrix multiplies is quite efficient while maintaining the effectiveness of the original, more-expensive forest model. Our neural approach has the advantage of being easier to train than any direct neural models, since it can match the previously-learned regression rather than learn to generalize relevance judgments directly. We observe CPU document scoring speed improvements of up to 400x over traditional algorithms and up to 10x over state-of-the-art algorithms with no measurable loss in mean average precision. With a GPU available, our algorithm is able to score every document in a batch in parallel for another 10-100x improvement. While we are not the first work to observe that neural networks are efficient as well as being effective, our application of this observation to learning to rank is novel and will have large real-world impact.

Unlike traditional learning to rank models that depend on hand-crafted features, neural representation learning models learn higher level features for the ranking task by training on large datasets. Their ability to learn new features directly from the data, however, may come at a price. Without any special supervision, these models learn relationships that may hold only in the domain from which the training data is sampled, and generalize poorly to domains not observed during training. We study the effectiveness of adversarial learning as a cross domain regularizer in the context of the ranking task. We use an adversarial discriminator and train our neural ranking model on a small set of domains. The discriminator provides a negative feedback signal to discourage the model from learning domain specific representations. Our experiments show consistently better performance on held out domains in the presence of the adversarial discriminator---sometimes up to 30% on [email protected]$.

With the rise in mobile and voice search, answer passage retrieval acts as a critical component of an effective information retrieval system for open domain question answering. Currently, there are no comparable collections that address non-factoid question answering within larger documents while simultaneously providing enough examples sufficient to train a deep neural network. In this paper, we introduce a new Wikipedia based collection specific for non-factoid answer passage retrieval containing thousands of questions with annotated answers and show benchmark results on a variety of state of the art neural architectures and retrieval models. The experimental results demonstrate the unique challenges presented by answer passage retrieval within topically relevant documents for future research.

Understanding and characterizing how people interact in information-seeking conversations is crucial in developing conversational search systems. In this paper, we introduce a new dataset designed for this purpose and use it to analyze information-seeking conversations by user intent distribution, co-occurrence, and flow patterns. The MSDialog dataset is a labeled dialog dataset of question answering (QA) interactions between information seekers and providers from an online forum on Microsoft products. The dataset contains more than 2,000 multi-turn QA dialogs with 10,000 utterances that are annotated with user intent on the utterance level. Annotations were done using crowdsourcing. With MSDialog, we find some highly recurring patterns in user intent during an information-seeking process. They could be useful for designing conversational search systems. We will make our dataset freely available to encourage exploration of information-seeking conversation models.

Intelligent personal assistant systems with either text-based or voice-based conversational interfaces are becoming increasingly popular around the world. Retrieval-based conversation models have the advantages of returning fluent and informative responses. Most existing studies in this area are on open domain ''chit-chat'' conversations or task / transaction oriented conversations. More research is needed for information-seeking conversations. There is also a lack of modeling external knowledge beyond the dialog utterances among current conversational models. In this paper, we propose a learning framework on the top of deep neural matching networks that leverages external knowledge for response ranking in information-seeking conversation systems. We incorporate external knowledge into deep neural models with pseudo-relevance feedback and QA correspondence knowledge distillation. Extensive experiments with three information-seeking conversation data sets including both open benchmarks and commercial data show that, our methods outperform various baseline methods including several deep text matching models and the state-of-the-art method on response selection in multi-turn conversations. We also perform analysis over different response types, model variations and ranking examples. Our models and research findings provide new insights on how to utilize external knowledge with deep neural models for response selection and have implications for the design of the next generation of information-seeking conversation systems.

Evidence derived from passages that closely represent likely answers to a posed query can be useful input to the ranking process. Based on a novel use of Community Question Answering data, we present an approach for the creation of such passages. A general framework for extracting answer passages and estimating their quality is proposed, and this evidence is integrated into ranking models. Our experiments on two web collections show that such quality estimates from answer passages provide a strong indication of document relevance and compare favorably to previous passage-based methods. Combining such evidence can significantly improve over a set of state-of-the-art ranking models, including Quality-Biased Ranking, External Expansion, and a combination of both. A final ranking model that incorporates all quality estimates achieves further improvements on both collections.

Predicting the performance of a search engine for a given query is a fundamental and challenging task in information retrieval. Accurate performance predictors can be used in various ways, such as triggering an action, choosing the most effective ranking function per query, or selecting the best variant from multiple query formulations. In this paper, we propose a general end-to-end query performance prediction framework based on neural networks, called NeuralQPP. Our framework consists of multiple components, each learning a representation suitable for performance prediction. These representations are then aggregated and fed into a prediction sub-network. We train our models with multiple weak supervision signals, which is an unsupervised learning approach that uses the existing unsupervised performance predictors using weak labels. We also propose a simple yet effective component dropout technique to regularize our model. Our experiments on four newswire and web collections demonstrate that NeuralQPP significantly outperforms state-of-the-art baselines, in nearly every case. Furthermore, we thoroughly analyze the effectiveness of each component, each weak supervision signal, and all resulting combinations in our experiments.

We formulate a document summarization method to extract passage-level answers for non-factoid queries, referred to as answer-biased summaries. We propose to use external information from related Community Question Answering (CQA) content to better identify answer bearing sentences. Three optimization-based methods are proposed: (i) query-biased, (ii) CQA-answer-biased, and (iii) expanded-query-biased, where expansion terms were derived from related CQA content. A learning-to-rank-based method is also proposed that incorporates a feature extracted from related CQA content. Our results show that even if a CQA answer does not contain a perfect answer to a query, their content can be exploited to improve the extraction of answer-biased summaries from other corpora. The quality of CQA content is found to impact on the accuracy of optimization-based summaries, though medium quality answers enable the system to achieve a comparable (and in some cases superior) accuracy to state-of-the-art techniques. The learning-to-rank-based summaries, on the other hand, are not significantly influenced by CQA quality. We provide a recommendation of the best use of our proposed approaches in regard to the availability of different quality levels of related CQA content. As a further investigation, the reliability of our approaches was tested on another publicly available dataset.

The Web has accumulated a rich source of information, such as text, image, rating, etc, which represent different aspects of user preferences. However, the heterogeneous nature of this information makes it difficult for recommender systems to leverage in a unified framework to boost the performance. Recently, the rapid development of representation learning techniques provides an approach to this problem. By translating the various information sources into a unified representation space, it becomes possible to integrate heterogeneous information for informed recommendation.

Product search is an important part of online shopping. In contrast to many search tasks, the objectives of product search are not confined to retrieving relevant products. Instead, it focuses on finding items that satisfy the needs of individuals and lead to a user purchase. The unique characteristics of product search make search personalization essential for both customers and e-shopping companies. Purchase behavior is highly personal in online shopping and users often provide rich feedback about their decisions (e.g. product reviews). However, the severe mismatch found in the language of queries, products and users make traditional retrieval models based on bag-of-words assumptions less suitable for personalization in product search. In this paper, we propose a hierarchical embedding model to learn semantic representations for entities (i.e. words, products, users and queries) from different levels with their associated language data. Our contributions are three-fold: (1) our work is one of the initial studies on personalized product search; (2) our hierarchical embedding model is the first latent space model that jointly learns distributed representations for queries, products and users with a deep neural network; (3) each component of our network is designed as a generative model so that the whole structure is explainable and extendable. Following the methodology of previous studies, we constructed personalized product search benchmarks with Amazon product data. Experiments show that our hierarchical embedding model significantly outperforms existing product search baselines on multiple benchmark datasets.

Incorporating conventional, unsupervised features into a neural architecture has the potential to improve modeling effectiveness, but this aspect is often overlooked in the research of deep learning models for information retrieval. We investigate this incorporation in the context of answer sentence selection, and show that combining a set of query matching, readability, and query focus features into a simple convolutional neural network can lead to markedly increased effectiveness. Our results on two standard question-answering datasets show the effectiveness of the combined model.

In recent years, deep neural networks have yielded significant performance improvements in application areas such as speech recognition, computer vision, and machine translation. This has led to expectations in the information retrieval (IR) community that these novel machine learning approaches are likely to demonstrate a similar scale of breakthroughs on IR tasks within the next couple of years. In the Neu-IR (pronounced "new IR") 2016 workshop, however, there was a growing concern that the lack of availability of large scale training and evaluation datasets may be hindering the research community from making adequate progress in this area. It was also highlighted that the community would benefit from establishing a shared public repository of neural IR models and shared evaluation resources for better reproducibility and speed of experimentation. After the first successful Neu-IR workshop at SIGIR 2016, our goal this year will be to host a highly interactive full-day workshop to bring the neural IR community together to specifically address these key challenges facing this line of research. The workshop will request the community to submit proposals on generating large scale benchmark collections, building a shared model repository, and standardizing frameworks appropriate for evaluating deep neural network models. In addition, the workshop will provide a forum for the growing community of IR researchers to present their recent (published and unpublished) work involving (shallow or deep) neural network based approaches in an interactive poster session.

Despite the impressive improvements achieved by unsupervised deep neural networks in computer vision and NLP tasks, such improvements have not yet been observed in ranking for information retrieval. The reason may be the complexity of the ranking problem, as it is not obvious how to learn from queries and documents when no supervised signal is available. Hence, in this paper, we propose to train a neural ranking model using weak supervision, where labels are obtained automatically without human annotators or any external resources (e.g., click data). To this aim, we use the output of an unsupervised ranking model, such as BM25, as a weak supervision signal. We further train a set of simple yet effective ranking models based on feed-forward neural networks. We study their effectiveness under various learning scenarios (point-wise and pair-wise models) and using different input representations (i.e., from encoding query-document pairs into dense/sparse vectors to using word embedding representation). We train our networks using tens of millions of training instances and evaluate it on two standard collections: a homogeneous news collection (Robust) and a heterogeneous large-scale web collection (ClueWeb). Our experiments indicate that employing proper objective functions and letting the networks to learn the input representation based on weakly supervised data leads to impressive performance, with over 13% and 35% MAP improvements over the BM25 model on the Robust and the ClueWeb collections. Our findings also suggest that supervised neural ranking models can greatly benefit from pre-training on large amounts of weakly labeled data that can be easily obtained from unsupervised IR models.

Learning a high-dimensional dense representation for vocabulary terms, also known as a word embedding, has recently attracted much attention in natural language processing and information retrieval tasks. The embedding vectors are typically learned based on term proximity in a large corpus. This means that the objective in well-known word embedding algorithms, e.g., word2vec, is to accurately predict adjacent word(s) for a given word or context. However, this objective is not necessarily equivalent to the goal of many information retrieval (IR) tasks. The primary objective in various IR tasks is to capture relevance instead of term proximity, syntactic, or even semantic similarity. This is the motivation for developing unsupervised relevance-based word embedding models that learn word representations based on query-document relevance information. In this paper, we propose two learning models with different objective functions; one learns a relevance distribution over the vocabulary set for each query, and the other classifies each term as belonging to the relevant or non-relevant class for each query. To train our models, we used over six million unique queries and the top ranked documents retrieved in response to each query, which are assumed to be relevant to the query. We extrinsically evaluate our learned word representation models using two IR tasks: query expansion and query classification. Both query expansion experiments on four TREC collections and query classification experiments on the KDD Cup 2005 dataset suggest that the relevance-based word embedding models significantly outperform state-of-the-art proximity-based embedding models, such as word2vec and GloVe.

We consider the problem of search result diversification for streams of short texts. Diversifying search results in short text streams is more challenging than in the case of long documents, as it is difficult to capture the latent topics of short documents. To capture the changes of topics and the probabilities of documents for a given query at a specific time in a short text stream, we propose a dynamic Dirichlet multinomial mixture topic model, called D2M3, as well as a Gibbs sampling algorithm for the inference. We also propose a streaming diversification algorithm, SDA, that integrates the information captured by D2M3 with our proposed modified version of the PM-2 (Proportionality-based diversification Method -- second version) diversification algorithm. We conduct experiments on a Twitter dataset and find that SDA statistically significantly outperforms state-of-the-art non-streaming retrieval methods, plain streaming retrieval methods, as well as streaming diversification methods that use other dynamic topic models.

In recent years, deep neural networks have led to exciting breakthroughs in speech recognition, computer vision, and natural language processing (NLP) tasks. However, there have been few positive results of deep models on ad-hoc retrieval tasks. This is partially due to the fact that many important characteristics of the ad-hoc retrieval task have not been well addressed in deep models yet. Typically, the ad-hoc retrieval task is formalized as a matching problem between two pieces of text in existing work using deep models, and treated equivalent to many NLP tasks such as paraphrase identification, question answering and automatic conversation. However, we argue that the ad-hoc retrieval task is mainly about relevance matching while most NLP matching tasks concern semantic matching, and there are some fundamental differences between these two matching tasks. Successful relevance matching requires proper handling of the exact matching signals, query term importance, and diverse matching requirements. In this paper, we propose a novel deep relevance matching model (DRMM) for ad-hoc retrieval. Specifically, our model employs a joint deep architecture at the query term level for relevance matching. By using matching histogram mapping, a feed forward matching network, and a term gating network, we can effectively deal with the three relevance matching factors mentioned above. Experimental results on two representative benchmark collections show that our model can significantly outperform some well-known retrieval models as well as state-of-the-art deep matching models.

A common limitation of many information retrieval (IR) models is that relevance scores are solely based on exact (i.e., syntactic) matching of words in queries and documents under the simple Bag-of-Words (BoW) representation. This not only leads to the well-known vocabulary mismatch problem, but also does not allow semantically related words to contribute to the relevance score. Recent advances in word embedding have shown that semantic representations for words can be efficiently learned by distributional models. A natural generalization is then to represent both queries and documents as Bag-of-Word-Embeddings (BoWE), which provides a better foundation for semantic matching than BoW. Based on this representation, we introduce a novel retrieval model by viewing the matching between queries and documents as a non-linear word transportation (NWT) problem. With this formulation, we define the capacity and profit of a transportation model designed for the IR task. We show that this transportation problem can be efficiently solved via pruning and indexing strategies. Experimental results on several representative benchmark datasets show that our model can outperform many state-of-the-art retrieval models as well as recently introduced word embedding-based models. We also conducted extensive experiments to analyze the effect of different settings on our semantic matching model.

Pseudo-relevance feedback (PRF) via query expansion has proven to be effective in many information retrieval tasks. In most existing work, the top-ranked documents from an initial search are assumed to be relevant and used for feedback. There are some drawbacks to this approach. One limitation is that there might be other relevant documents which were not retrieved or considered for the the feedback process. Another issue is one or more of the top retrieved documents may be non-relevant, which can introduce noise into the feedback mechanism. Term-level diversification, on the other hand, uses an effective technique for identifying terms associated with query aspects or subtopics. We propose a new iterative feedback method that combines PRF with aspect generation to improve feedback effectiveness. In our experiments, we discovered a new property of convergence of feedback terms that was incorporated into the PRF process. We show that the resulting method significantly outperforms the baseline relevance model.

As an alternative to question answering methods based on feature engineering, deep learning approaches such as convolutional neural networks (CNNs) and Long Short-Term Memory Models (LSTMs) have recently been proposed for semantic matching of questions and answers. To achieve good results, however, these models have been combined with additional features such as word overlap or BM25 scores. Without this combination, these models perform significantly worse than methods based on linguistic feature engineering. In this paper, we propose an attention based neural matching model for ranking short answer text. We adopt value-shared weighting scheme instead of position-shared weighting scheme for combining different matching signals and incorporate question term importance learning using question attention network. Using the popular benchmark TREC QA data, we show that the relatively simple aNMM model can significantly outperform other neural network models that have been used for the question answering task, and is competitive with models that are combined with additional features. When aNMM is combined with additional features, it outperforms all baselines.

In information retrieval, pseudo-relevance feedback (PRF) refers to a strategy for updating the query model using the top retrieved documents. PRF has been proven to be highly effective in improving the retrieval performance. In this paper, we look at the PRF task as a recommendation problem: the goal is to recommend a number of terms for a given query along with weights, such that the final weights of terms in the updated query model better reflect the terms' contributions in the query. To do so, we propose RFMF, a PRF framework based on matrix factorization which is a state-of-the-art technique in collaborative recommender systems. Our purpose is to predict the weight of terms that have not appeared in the query and matrix factorization techniques are used to predict these weights. In RFMF, we first create a matrix whose elements are computed using a weight function that shows how much a term discriminates the query or the top retrieved documents from the collection. Then, we re-estimate the created matrix using a matrix factorization technique. Finally, the query model is updated using the re-estimated matrix. RFMF is a general framework that can be employed with any retrieval model. In this paper, we implement this framework for two widely used document retrieval frameworks: language modeling and the vector space model. Extensive experiments over several TREC collections demonstrate that the RFMF framework significantly outperforms competitive baselines. These results indicate the potential of using other recommendation techniques in this task.

Incorporating topic level estimation into language models has been shown to be beneficial for information retrieval (IR) models such as cluster-based retrieval and LDA-based document representation. Neural embedding models, such as paragraph vector (PV) models, on the other hand have shown their effectiveness and efficiency in learning semantic representations of documents and words in multiple Natural Language Processing (NLP) tasks. However, their effectiveness in information retrieval is mostly unknown. In this paper, we study how to effectively use the PV model to improve ad-hoc retrieval. We propose three major improvements over the original PV model to adapt it for the IR scenario: (1) we use a document frequency-based rather than the corpus frequency-based negative sampling strategy so that the importance of frequent words will not be suppressed excessively; (2) we introduce regularization over the document representation to prevent the model overfitting short documents along with the learning iterations; and (3) we employ a joint learning objective which considers both the document-word and word-context associations to produce better word probability estimation. By incorporating this enhanced PV model into the language modeling framework, we show that it can significantly outperform the state-of-the-art topic enhanced language models.

In recent years, deep neural networks have yielded significant performance improvements on speech recognition and computer vision tasks, as well as led to exciting breakthroughs in novel application areas such as automatic voice translation, image captioning, and conversational agents. Despite demonstrating good performance on natural language processing (NLP) tasks (e.g., language modelling and machine translation, the performance of deep neural networks on information retrieval (IR) tasks has had relatively less scrutiny. Recent work in this area has mainly focused on word embeddings and neural models for short text similarity.

Developing effective methods for fusing multiple ranked lists of documents is crucial to many applications. Federated web search, for instance, has become a common practice where a query is issued to different verticals and a single ranked list of blended results is created. While federated search is regarded as collection fusion, data fusion techniques aim at improving search coverage and precision by combining multiple search runs on a single document collection. In this paper, we study in depth and extend a neural network-based approach, LambdaMerge, for merging results of ranked lists drawn from one (i.e., data fusion) or more (i.e., collection fusion) verticals. The proposed model considers the impact of the quality of documents, ranked lists and verticals for producing the final merged result in an optimization framework. We further investigate the potential of incorporating deep structures into the model with an aim of determining better combinations of different evidence. In the experiments on collection fusion and data fusion, the proposed approach significantly outperforms several standard baselines and state-of-the-art learning-based approaches.

We demonstrate an exploration tool that organizes social media content under diverse aspects enabling comprehensive explorations. Unlike existing approaches that group content by trending topics, we present a holistic view of diverse and relevant content with respect to a given query.

Many queries, especially those in the form of longer questions, contain a subset of terms representing key concepts that describe the most important part of the user's information need. Detecting the key concepts in a query can be used as the basis for more effective weighting of query terms, but in this paper, we focus on a method of using the key concepts in a translation model for query expansion and retrieval. Translation models have been used previously in community-based question answering (CQA) systems in order to bridge the semantic gap between questions and the corresponding answer documents. Our method uses the key concepts of a question as the translation context and selectively applies the translation model to the secondary (non-key) parts of the question. We evaluate the proposed method using a CQA collection and show that selectively translating key and secondary concepts can significantly improve the retrieval performance compared to a baseline that applies the translation model without considering key concepts.

A number of retrieval models incorporating term dependencies have recently been introduced. Most of these modify existing "bag-of-words" retrieval models by including features based on the proximity of pairs of terms (or bi-terms). Although these term dependency models have been shown to be significantly more effective than the bag-of-words models, there have been no previous systematic comparisons between the different approaches that have been proposed. In this paper, we compare the effectiveness of recent bi-term dependency models over a range of TREC collections, for both short (title) and long (description) queries. To ensure the reproducibility of our study, all experiments are performed on widely available TREC collections, and all tuned retrieval model parameters are made public. These comparisons show that the weighted sequential dependence model is at least as effective as, and often significantly better than, any other model across this range of collections and queries. We observe that dependency features are much more valuable in improving the performance of longer queries than for shorter queries. We then examine the effectiveness of dependence models that incorporate proximity features involving more than two terms. The results show that these features can improve effectiveness, but not consistently, over the available data sets.

Relevance feedback has been shown to improve retrieval for a broad range of retrieval models. It is the most common way of adapting a retrieval model for a specific query. In this work, we expand this common way by focusing on an approach that enables us to do query-specific modification of a retrieval model for learning-to-rank problems. Our approach is based on using feedback documents in two ways: 1) to improve the retrieval model directly and 2) to identify a subset of training queries that are more predictive than others. Experiments with the Gov2 collection show that this approach can obtain statistically significant improvements over two baselines; learning-to-rank (SVM-rank) with no feedback and learning-to-rank with standard relevance feedback.

Passage-based retrieval models have been studied for some time and have been shown to have some benefits for document ranking. Finding passages that are not only topically relevant, but are also answers to the users' questions would have a significant impact in applications such as mobile search. To develop models for answer passage retrieval, we need to have appropriate test collections and evaluation measures. Making annotations at the passage level is, however, expensive and can have poor coverage. In this paper, we describe the advantages of document summarization measures for evaluating answer passage retrieval and show that these measures have high correlation with existing measures and human judgments.

Many domain-specific search tasks are initiated by document-length queries, e.g., patent invalidity search aims to find prior art related to a new (query) patent. We call this type of search Query Document Search. In this type of search, the initial query document is typically long and contains diverse aspects (or sub-topics). Users tend to issue many queries based on the initial document to retrieve relevant documents. To help users in this situation, we propose a method to suggest diverse queries that can cover multiple aspects of the query document. We first identify multiple query aspects and then provide diverse query suggestions that are effective for retrieving relevant documents as well being related to more query aspects. In the experiments, we demonstrate that our approach is effective in comparison to previous query suggestion methods.

Formulating and processing phrases and other term dependencies to improve query effectiveness is an important problem in information retrieval. However, accessing word-sequence statistics using inverted indexes requires unreasonable processing time or substantial space overhead. Establishing a balance between these competing space and time trade-offs can dramatically improve system performance. In this article, we present and analyze a new index structure designed to improve query efficiency in dependency retrieval models. By adapting a class of ( ε, δ )-approximation algorithms originally proposed for sketch summarization in networking applications, we show how to accurately estimate statistics important in term-dependency models with low, probabilistically bounded error rates. The space requirements for the vocabulary of the index is only logarithmically linked to the size of the vocabulary. Empirically, we show that the sketch index can reduce the space requirements of the vocabulary component of an index of n -grams consisting of between 1 and 4 words extracted from the GOV2 collection to less than 0.01% of the space requirements of the vocabulary of a full index. We also show that larger n -gram queries can be processed considerably more efficiently than in current alternatives, such as positional and next-word indexes.

Abstract Understanding the landscape of opinions on a given topic or issue is important for policy makers, sociologists, and intelligence analysts. The first step in this process is to retrieve relevant opinions. Discussion forums are potentially a good source of this information, but comes with a unique set of retrieval challenges. In this short paper, we test a range of existing techniques for forum retrieval and develop new retrieval models to differentiate between opinionated and factual forum posts. We are able to demonstrate some significant performance improvements over the baseline retrieval models, demonstrating that this as a promising avenue for further study.

The goal of a blog retrieval system is to retrieve and rank blogs, as collections of documents, in response to a given query. Previous studies have shown that diversity among the top retrieved posts from a blog is a positive feature for indicating relevance of the blog to the query. However, existing methods capture the diversity of a blog using post-level properties that limits their application to a specific category of retrieval methods. In this paper, we propose a blog-level diversity measure where there is no assumption made about the underlying blog-ranking technique. The proposed measure enables us to integrate diversity in any existing blog retrieval method. Our experimental results show that the proposed method, while being more general, produces comparable results to the post-level diversity detection methods.

Current approaches for search result diversification have been categorized as either implicit or explicit. The implicit approach assumes each document represents its own topic, and promotes diversity by selecting documents for different topics based on the difference of their vocabulary. On the other hand, the explicit approach models the set of query topics, or aspects. While the former approach is generally less effective, the latter usually depends on a manually created description of the query aspects, the automatic construction of which has proven difficult. This paper introduces a new approach: term-level diversification. Instead of modeling the set of query aspects, which are typically represented as coherent groups of terms, our approach uses terms without the grouping. Our results on the ClueWeb collection show that the grouping of topic terms provides very little benefit to diversification compared to simply using the terms themselves. Consequently, we demonstrate that term-level diversification, with topic terms identified automatically from the search results using a simple greedy algorithm, significantly outperforms methods that attempt to create a full topic structure for diversification.

Community Question Answering (CQA) platforms contain a large number of questions and associated answers. Answerers sometimes include URLs as part of the answers to provide further information. This paper describes a novel way of building a test collection for web search by exploiting the link information from this type of social media data. We propose to build the test collection by regarding CQA questions as queries and the associated linked web pages as relevant documents. To evaluate this approach, we collect approximately ten thousand CQA queries, whose answers contained links to ClueWeb09 documents after spam filtering. Experimental results using this collection show that the relative effectiveness between different retrieval models on the ClueWeb-CQA query set is consistent with that on the TREC Web Track query sets, confirming the reliability of our test collection. Further analysis shows that the large number of queries generated through this approach compensates for the sparse relevance judgments in determining significant differences.

Many recent and highly effective retrieval models for long queries use query reformulation methods that jointly optimize term weights and term selection. These methods learn using word context and global context but typically fail to capture query context. In this paper, we present a novel term ranking algorithm, PhRank, that extends work on Markov chain frameworks for query expansion to select compact and focused terms from within a query itself. This focuses queries so that one to five terms in an unweighted model achieve better retrieval effectiveness than weighted term selection models that use up to 30 terms. PhRank terms are also typically compact and contain 1-2 words compared to competing models that use query subsets up to 7 words long. PhRank captures query context with an affinity graph constructed using word co-occurrence in pseudo-relevant documents. A random walk of the graph is used for term ranking in combination with discrimination weights. Empirical evaluation using newswire and web collections demonstrates that performance of reformulated queries is significantly improved for long queries and at least as good for short, keyword queries compared to highly competitive information retrieval (IR) models.

Query reformulation modifies the original query with the aim of better matching the vocabulary of the relevant documents, and consequently improving ranking effectiveness. Previous models typically generate words and phrases related to the original query, but do not consider how these words and phrases would fit together in actual queries. In this article, a novel framework is proposed that models reformulation as a distribution of actual queries, where each query is a variation of the original query. This approach considers an actual query as the basic unit and thus captures important query-level dependencies between words and phrases. An implementation of this framework that only uses publicly available resources is proposed, which makes fair comparisons with other methods using TREC collections possible. Specifically, this implementation consists of a query generation step that analyzes the passages containing query words to generate reformulated queries and a probability estimation step that learns a distribution for reformulated queries by optimizing the retrieval performance. Experiments on TREC collections show that the proposed model can significantly outperform previous reformulation models.

Time information impacts relevance in retrieval for the queries that are sensitive to trends and events. Microblog services particularly focused on recent news and events so dealing with the temporal aspects of microblogs is essential for providing effective retrieval. Recent work on time-based retrieval has shown that selecting the relevant time period for query expansion is promising. In this paper, we suggest a method for selecting the time period for query expansion based on a user behavior (i.e., retweets) that can be collected easily. We then use these time periods for query expansion in a pseudo-relevance feedback setting. More specifically, we use the difference in the temporal distribution between the top retrieved documents and retweets. The experimental results based on the TREC Microblog collection show that our method for selecting periods for query expansion improves retrieval performance compared to another approach.

Microblog services typically contain very short documents (e.g., tweets) containing comments about the latest news and events. Many of these documents are not informative or have very little content due to their personal and ephemeral nature. Providing effective retrieval in a microblog service will require addressing the challenge of distinguishing the high-quality, informative documents from the others. Recent work has focused on finding features that indicate the quality of microblog documents, but the impact these quality features on retrieval is not clear. In this paper, we suggest a low-cost quality model using surrogate judgments based on user behavior (i.e., retweets) that can be collected automatically. We analyze the relationship between document informativeness and relevance judgments for microblog retrieval. Then we demonstrate that our behavior-based quality metric has a high correlation with manual judgments. Also, we perform experiments to study the impact of the quality model on microblog retrieval. The results based on the TREC Microblog track show that the proposed quality model, combined with a variety of retrieval models, can improve retrieval performance and is competitive with a model trained using manual relevance judgments.

Formulating and processing phrases and other term dependencies to improve query effectiveness is an important problem in information retrieval. However, accessing these types of statistics using standard inverted indexes requires unreasonable processing time or incurs a substantial space overhead. Establishing a balance between these competing space and time trade-offs can dramatically improve system performance.

Blog distillation (blog feed retrieval) is a task in blog retrieval where the goal is to rank blogs according to their recurrent relevance to a query topic. One of the main properties of blog feed retrieval is that the unit of retrieval is a collection of documents as opposed to a single document as in other IR tasks. This collection retrieval nature of blog distillation introduces new challenges and requires new investigations specific to this problem.

Many of the recent, and more effective, retrieval models have incorporated dependencies between the terms in the query. In this paper, we advance this query representation one step further, and propose a retrieval framework that models higher-order term dependencies, i.e., dependencies between arbitrary query concepts rather than just query terms. In order to model higher-order term dependencies, we represent a query using a hypergraph structure -- a generalization of a graph, where a (hyper)edge connects an arbitrary subset of vertices. A vertex in a query hypergraph corresponds to an individual query concept, and a dependency between a subset of these vertices is modeled through a hyperedge. An extensive empirical evaluation using both newswire and web corpora demonstrates that query representation using hypergraphs is highly beneficial for verbose natural language queries. For these queries, query hypergraphs significantly improve the retrieval effectiveness of several state-of-the-art models that do not employ higher-order term dependencies.

We propose a kernel-based model to automatically extract social relations such as economic relations and political relations between two people from news articles. To determine whether two people are structurally associated with each other, the proposed model uses an SVM (support vector machine) tree kernel based on trigrams of head-dependent relations between them. In the experiments with the automatic content extraction (ACE) corpus and a Korean news corpus, the proposed model outperformed the previous systems based on SVM tree kernels even though it used more shallow linguistic knowledge.

This paper presents a different perspective on diversity in search results: diversity by proportionality. We consider a result list most diverse, with respect to some set of topics related to the query, when the number of documents it provides on each topic is proportional to the topic's popularity. Consequently, we propose a framework for optimizing proportionality for search result diversification, which is motivated by the problem of assigning seats to members of competing political parties. Our technique iteratively determines, for each position in the result ranked list, the topic that best maintains the overall proportionality. It then selects the best document on this topic for this position. We demonstrate empirically that our method significantly outperforms the top performing approach in the literature not only on our proposed metric for proportionality, but also on several standard diversity measures. This result indicates that promoting proportionality naturally leads to minimal redundancy, which is a goal of the current diversity approaches.

Search queries have evolved beyond keyword queries. Many complex queries such as verbose queries, natural language question queries and document-based queries are widely used in a variety of applications. Processing these complex queries usually requires a series of query operations, which results in multiple sequences of reformulated queries. However, previous query representations, either the "bag of words" method or the recently proposed "query distribution" method, cannot effectively model these query sequences, since they ignore the relationships between two queries. In this paper, a reformulation tree framework is proposed to organize multiple sequences of reformulated queries as a tree structure, where each path of the tree corresponds to a sequence of reformulated queries. Specifically, a two-level reformulation tree is implemented for verbose queries. This tree effectively combines two query operations, i.e., subset selection and query substitution, within the same framework. Furthermore, a weight estimation approach is proposed to assign weights to each node of the reformulation tree by considering the relationships with other nodes and directly optimizing retrieval performance. Experiments on TREC collections show that this reformulation tree based representation significantly outperforms the state-of-the-art techniques.

Most standard information retrieval models use a single source of information (e.g., the retrieval corpus) for query formulation tasks such as term and phrase weighting and query expansion. In contrast, in this paper, we present a unified framework that automatically optimizes the combination of information sources used for effective query formulation. The proposed framework produces fully weighted and expanded queries that are both more effective and more compact than those produced by the current state-of-the-art query expansion and weighting methods. We conduct an empirical evaluation of our framework for both newswire and web corpora. In all cases, our combination of multiple information sources for query formulation is found to be more effective than using any single source. The proposed query formulations are especially advantageous for large scale web corpora, where they also reduce the number of terms required for effective query expansion, and improve the diversity of the retrieved results.

The prevalence of social media applications is generating potentially large personal archives of posts, tweets, and other communications. The existence of these archives creates a need for search tools, which can be seen as an extension of current desktop search services. Little is currently known about the best search techniques for personal archives of social data, because of the difficulty of creating test collections. In this paper, we describe how test collections for personal social data can be created by using games to collect queries. We then compare a range of retrieval models that exploit the semi-structured nature of social data. Our results show that a mixture of language models with field distribution estimation can be effective for this type of data, with certain fields, such as the name of the poster, being particularly important. We also analyze the properties of the queries that were generated by users with two versions of the games.

When the information need is not clear from the user query, a good strategy would be to return documents that cover as many aspects of the query as possible. To do this, the possible aspects of the query need to be automatically identified. In this paper, we propose to do this by clustering reformulated queries generated from publicly available resources and using each cluster to represent an aspect of the query. Our results show that the automatically generated reformulations for the TREC Web Track queries match up quite well with actual sub-topics of these queries identified by TREC experts. Moreover, agglomerative clustering using query-to-query similarity based on co-occurrence in text passages can provide clusters of high quality that potentially can be used to identify aspects.

Pseudo relevance feedback (PRF) is one of effective practices in Information Retrieval. In particular, PRF via the relevance model (RM) has been widely used due to the theoretical soundness and effectiveness. In a PRF scenario, an underlying relevance model is inferred by combining language models of the top retrieved documents where the contribution of each document is assumed to be proportional to its score for the initial query. However, it is not clear that selecting the top retrieved documents only by the initial retrieval scores is actually the optimal way for query expansion.

Recent studies suggest that associative browsing can be beneficial for personal information access. Associative browsing is intuitive for the user and complements other methods of accessing personal information, such as keyword search. In our previous work, we proposed an associative browsing model of personal information in which users can navigate through the space of documents and concepts (e.g., person names, events, etc.). Our approach differs from other systems in that it presented a ranked list of associations by combining multiple measures of similarity, whose weights are improved based on click feedback from the user.

Incorporating syntactic features in a retrieval model has had very limited success in the past, with the exception of binary term dependencies. This paper presents a new term dependency modeling approach based on syntactic dependency parsing for both queries and documents. Our model is inspired by a quasi-synchronous stochastic process for machine translation[21]. We model four different types of relationships between syntactically dependent term pairs to perform inexact matching between documents and queries. We also propose a machine learning technique for predicting optimal parameter settings for a retrieval model incorporating syntactic relationships. The results on TREC collections show that the quasi-synchronous dependence model can improve retrieval performance and outperform a strong state-of-art sequential dependence baseline when we use predicted optimal parameters.

The increasing volume of streaming data on microblogs has re-introduced the necessity of effective filtering mechanisms for such media. Microblog users are overwhelmed with mostly uninteresting pieces of text in order to access information of value. In this paper, we propose a personalized tweet ranking method, leveraging the use of retweet behavior, to bring more important tweets forward. In addition, we also investigate how to determine the audience of tweets more effectively, by ranking the users based on their likelihood of retweeting the tweets. Finally, conducting a pilot user study, we analyze how retweet likelihood correlates with the interestingness of the tweets.

The majority of the current information retrieval models weight the query concepts (e.g., terms or phrases) in an unsupervised manner, based solely on the collection statistics. In this paper, we go beyond the unsupervised estimation of concept weights, and propose a parameterized concept weighting model. In our model, the weight of each query concept is determined using a parameterized combination of diverse importance features. Unlike the existing supervised ranking methods, our model learns importance weights not only for the explicit query concepts, but also for the latent concepts that are associated with the query through pseudo-relevance feedback. The experimental results on both newswire and web TREC corpora show that our model consistently and significantly outperforms a wide range of state-of-the-art retrieval models. In addition, our model significantly reduces the number of latent concepts used for query expansion compared to the non-parameterized pseudo-relevance feedback based models.

In professional search environments, such as patent search or legal search, search tasks have unique characteristics: 1) users interactively issue several queries for a topic, and 2) users are willing to examine many retrieval results, i.e., there is typically an emphasis on recall. Recent surveys have also verified that professional searchers continue to have a strong preference for Boolean queries because they provide a record of what documents were searched. To support this type of professional search, we propose a novel Boolean query suggestion technique. Specifically, we generate Boolean queries by exploiting decision trees learned from pseudo-labeled documents and rank the suggested queries using query quality predictors. We evaluate our algorithm in simulated patent and medical search environments. Compared with a recent effective query generation system, we demonstrate that our technique is effective and general.

Improving verbose (or long) queries poses a new challenge for search systems. Previous techniques mainly focused on two aspects, weighting the important words or phrases and selecting the best subset query. The former does not consider how words and phrases are used in actual subset queries, while the latter ignores alternative subset queries. Recently, a novel reformulation framework has been proposed to transform the original query as a distribution of reformulated queries, which overcomes the disadvantages of previous techniques. In this paper, we apply this framework to verbose queries, where a reformulated query is specified as a subset query. Experiments on TREC collections show that the query distribution based framework outperforms the state-of-the-art techniques.

Many existing retrieval approaches do not take into account the content quality of the retrieved documents, although link-based measures such as PageRank are commonly used as a form of document prior. In this paper, we present the quality-biased ranking method that promotes documents containing high-quality content, and penalizes low-quality documents. The quality of the document content can be determined by its readability, layout and ease-of-navigation, among other factors. Accordingly, instead of using a single estimate for document quality, we consider multiple content-based features that are directly integrated into a state-of-the-art retrieval method. These content-based features are easy to compute, store and retrieve, even for large web col-lections. We use several query sets and web collections to empirically evaluate the performance of our quality-biased retrieval method. In each case, our method consistently im-proves by a large margin the retrieval performance of text-based and link-based retrieval methods that do not take into account the quality of the document content.

The identification of repeated n-gram phrases in text has many practical applications, including authorship attribution, text reuse identification, and plagiarism detection. We consider methods for finding the repeated n-grams in text corpora, with emphasis on techniques that can be effectively scaled across a cluster of processors to handle very large amounts of text. We compare our proposed method to existing techniques using the 1.5 TB TREC ClueWeb-B text collection, using both single-processor and multi-processor approaches. The experiments show that our method offers an important tradeoff between speed and temporary storage space, and provides an alternative to previous approaches that scales almost linearly in the length of the sequence, is largely independent of n, and provides a uniform workload balance across the set of available processors.

Marking up queries with annotations such as part-of-speech tags, capitalization, and segmentation, is an important part of many approaches to query processing and understanding. Due to their brevity and idiosyncratic structure, search queries pose a challenge to existing annotation tools that are commonly trained on full-length documents. To address this challenge, we view the query as an explicit representation of a latent information need, which allows us to use pseudo-relevance feedback, and to leverage additional information from the document corpus, in order to improve the quality of query annotation.

A typical collection of personal information contains many documents and mentions many concepts (e.g., person names, events, etc.). In this environment, associative browsing between these concepts and documents can be useful as a complement for search. Previous approaches in the area of semantic desktops aimed at addressing this task. However, they were not practical because they require tedious manual annotation by the user.

Query reformulation modifies the original query with the aim of better matching the vocabulary of the relevant documents, and consequently improving ranking effectiveness. Previous techniques typically generate words and phrases related to the original query, but do not consider how these words and phrases would fit together in new queries. In this paper, we focus on an implementation of an approach that models reformulation as a distribution of queries, where each query is a variation of the original query. This approach considers a query as a basic unit and can capture important dependencies between words and phrases in the query. The implementation discussed here is based on passage analysis of the target corpus. Experiments on the TREC collection show that the proposed model for query reformulation significantly outperforms state-of-the-art methods.

Dealing with verbose (or long) queries poses a new challenge for information retrieval. Selecting a subset of the original query (a "sub-query") has been shown to be an effective method for improving these queries. In this paper, the distribution of sub-queries ("subset distribution") is formally modeled within a well-grounded framework. Specifically, sub-query selection is considered as a sequential labeling problem, where each query word in a verbose query is assigned a label of "keep" or "don't keep". A novel Conditional Random Field model is proposed to generate the distribution of sub-queries. This model captures the local and global dependencies between query words and directly optimizes the expected retrieval performance on a training set. The experiments, based on different retrieval models and performance measures, show that the proposed model can generate high-quality sub-query distributions and can significantly outperform state-of-the-art techniques.

Query reformulation techniques based on query logs have recently proven to be effective for web queries. However, when initial queries have reasonably good quality, these techniques are often not reliable enough to identify the helpful reformulations among the suggested queries. In this paper, we show that we can use as few as two features to rerank a list of reformulated queries, or expanded queries to be specific, generated by a log-based query reformulation technique. Our results across five TREC collections suggest that there are consistently more useful reformulations in the first two positions in the new ranked list than there were initially, which leads to statistically significant improvements in retrieval effectiveness.

Verbose or long queries are a small but significant part of the query stream in web search, and are common in other applications such as collaborative question answering (CQA). Current search engines perform well with keyword queries but are not, in general, effective for verbose queries. In this paper, we examine query processing techniques which can be applied to verbose queries prior to submission to a search engine in order to improve the search engine's results. We focus on verbose queries that have sentence-like structure, but are not simple "wh-" questions, and assume the search engine is a "black box." We evaluated the output of two search engines using queries from a CQA service and our results show that, among a broad range of techniques, the most effective approach is to simply reduce the length of the query. This can be achieved effectively by removing "stop structure" instead of only stop words. We show that the process of learning and removing stop structure from a query can be effectively automated.

A typical desktop environment contains many document types (email, presentations, web pages, pdfs, etc.) each with different metadata. Predicting which types of documents a user is looking for in the context of a given query is a crucial part of providing effective desktop search. The problem is similar to selecting resources in distributed IR, but there are some important differences.

Query term ranking approaches are used to select effective terms from a verbose query by ranking terms. Features used for query term ranking and selection in previous work do not consider grammatical relationships between terms. To address this issue, we use syntactic features extracted from dependency parsing results of verbose queries. We also modify the method for measuring the effectiveness of query terms for query term ranking.

Combining multiple documents to represent an information object is well-known as an effective approach for many Information Retrieval tasks. For example, passages can be combined to represent a document for retrieval, document clusters are represented using combinations of the documents they contain, and feedback documents can be combined to represent a query model. Various techniques for combination have been introduced, and among them, representation techniques based on concatenation and the arithmetic mean are frequently used. Some recent work has shown the potential of a new representation technique using the geometric mean. However, these studies lack a theoretical foundation explaining why the geometric mean should have advantages for representing multiple documents. In this paper, we show that the arithmetic mean and the geometric mean are approximations to the center of mass in certain geometries, and show empirically that the geometric mean is closer to the center. Through experiments with two IR tasks, we show the potential benefits for geometric representations, including a geometry-based pseudo-relevance feedback method that outperforms state-of-the-art techniques.

A standard approach for determining a Dirichlet smoothing parameter is to choose a value which maximizes a retrieval performance metric using training data consisting of queries and relevance judgments. There are, however, situations where training data does not exist or the queries and relevance judgments do not reflect typical user information needs for the application. We propose an unsupervised approach for estimating a Dirichlet smoothing parameter based on collection statistics. We show empirically that this approach can suggest a plausible Dirichlet smoothing parameter value in cases where relevance judgments cannot be used.

Modeling query concepts through term dependencies has been shown to have a significant positive effect on retrieval performance, especially for tasks such as web search, where relevance at high ranks is particularly critical. Most previous work, however, treats all concepts as equally important, an assumption that often does not hold, especially for longer, more complex queries. In this paper, we show that one of the most effective existing term dependence models can be naturally extended by assigning weights to concepts. We demonstrate that the weighted dependence model can be trained using existing learning-to-rank techniques, even with a relatively small number of training queries. Our study compares the effectiveness of both endogenous (collection-based) and exogenous (based on external sources) features for determining concept importance. To test the weighted dependence model, we perform experiments on both publicly available TREC corpora and a proprietary web corpus. Our experimental results indicate that our model consistently and significantly outperforms both the standard bag-of-words model and the unweighted term dependence model, and that combining endogenous and exogenous features generally results in the best retrieval effectiveness.

Query reformulation techniques based on query logs have been studied as a method of capturing user intent and improving retrieval effectiveness. The evaluation of these techniques has primarily, however, focused on proprietary query logs and selected samples of queries. In this paper, we suggest that anchor text, which is readily available, can be an effective substitute for a query log and study the effectiveness of a range of query reformulation techniques (including log-based stemming, substitution, and expansion) using standard TREC collections. Our results show that log-based query reformulation techniques are indeed effective with standard collections, but expansion is a much safer form of query modification than word substitution. We also show that using anchor text as a simulated query log is as least as effective as a real log for these techniques.

Representing the information need is the greatest challenge for opinion retrieval. Typical queries for opinion retrieval are composed of either just content words, or content words with a small number of cue "opinion" words. Both are inadequate for retrieving opinionated documents. In this paper, we develop a general formal framework--the opinion relevance model--to represent an information need for opinion retrieval. We explore a series of methods to automatically identify the most appropriate opinion words for query expansion, including using query independent sentiment resources. We also propose a relevance feedback-based approach to extract opinion words. Both query-independent and query-dependent methods can also be integrated into a more effective mixture relevance model. Finally, opinion retrieval experiments are presented for the Blog06 and COAE08 text collections. The results show that, significant improvements can always be obtained by this opinion relevance model whether sentiment resources are available or not.

Desktop search is an important part of personal information management (PIM). However, research in this area has been limited by the lack of shareable test collections, making cumulative progress difficult. In this paper, we define desktop search as a semi-structured document retrieval problem and introduce a methodology to automatically build a reusable collection (the pseudo-desktop) that has many of the same properties as a real desktop collection. We then present a comprehensive evaluation of retrieval methods for semi-structured document retrieval on several pseudo-desktop collections and the TREC Enterprise collection. Our results show that a probabilistic retrieval model using the mapping relation between a query term and a document field (PRM-S) has the best performance in collections with more structure, such as email, and that the query-likelihood language model is better for other document types. We further analyze the observed differences using generated queries and suggest ways to improve PRM-S, which makes the performance gains more significant and consistent.

Online communities are valuable information sources where knowledge is accumulated by interactions between people. Search services provided by online community sites such as forums are often, however, quite poor. To address this, we investigate retrieval techniques that exploit the hierarchical thread structures in community sites. Since these structures are sometimes not explicit or accurately annotated, we use structure discovery techniques. We then make use of thread structures in retrieval experiments. Our results show that using thread structures that have been accurately annotated can lead to significant improvements in retrieval performance compared to strong baselines.

Patent search is the task of finding relevant existing patents, which is an important part of the patent's examiner's process of validating a patent application. In this paper, we studied how to transform a query patent (the application) into search queries. Three types of search features are explored for automatic query generation for patent search. Furthermore, different types of features are combined with a learning to rank method. Experiments based on a USPTO patent collection demonstrate that the single best search feature is the combination of words and noun-phrases from the summary field and the retrieval performance can be significantly improved by combining three types of search features.

Modeling term dependence has been shown to have a significant positive impact on retrieval. Current models, however, use sequential term dependencies, leading to an increased query latency, especially for long queries. In this paper, we examine two query segmentation models that reduce the number of dependencies. We find that two-stage segmentation based on both query syntactic structure and external information sources such as query logs, attains retrieval performance comparable to the sequential dependence model, while achieving a 50% reduction in query latency.

Searching for prior-art patents is an essential step for the patent examiner to validate or invalidate a patent application. In this paper, we consider the whole patent as the query, which reduces the burden on the user, and also makes many more potential search features available. We explore how to automatically transform the query patent into an effective search query, especially focusing on the effect of different patent fields. Experiments show that the background summary of a patent is the most useful source of terms for generating a query, even though most previous work used the patent claims.

With the overwhelming number of reports on similar events originating from different sources on the web, it is often hard, using existing web search paradigms, to find the original source of 'facts', statements, rumors, and opinions, and to track their development. Several techniques have been previously proposed for detecting such text reuse between different sources, however these techniques have been tested against relatively small and homogeneous TREC collections. In this work, we test the feasibility of text reuse detection techniques in the setting of web search. In addition to text reuse detection, we develop a novel technique that addresses the unique challenges of finding original sources on the web, such as defining a timeline. We also explore the use of link analysis for identifying reliable and relevant reports. Our experimental results show that the proposed techniques can operate on the scale of the web, are significantly more accurate than standard web search for finding text reuse, and provide a richer representation for tracking the information flow.

Search applications have become ubiquitous and very successful. Major advances have been made in understanding how to deliver effective results very efficiently for a class of queries. As the range of applications broaden to include web search, desktop search, enterprise search, vertical search, social search, etc., the number of new research challenges has appeared to grow rather than shrink. Many of these challenges are variations on underlying themes and principles that information retrieval has focused on for more than 40 years. In this talk, several unsolved problems arising from new search applications will be discussed and some potential paths to solutions for these problems will be outlined.

A blog site consists of many individual blog postings. Current blog search services focus on retrieving postings but there is also a need to identify relevant blog sites. Blog site search is similar to resource selection in distributed information retrieval, in that the target is to find relevant collections of documents. We introduce resource selection techniques for blog site search and evaluate their performance. Further, we propose a "diversity factor" that measures the topic diversity of each blog site. Our results show that the appropriate combination of the resource selection techniques and the diversity factor can achieve significant improvements in retrieval performance compared to baselines. We also report results using these techniques on the TREC blog distillation task.

Current search engines do not, in general, perform well with longer, more verbose queries. One of the main issues in processing these queries is identifying the key concepts that will have the most impact on effectiveness. In this paper, we develop and evaluate a technique that uses query-dependent, corpus-dependent, and corpus-independent features for automatic extraction of key concepts from verbose queries. We show that our method achieves higher accuracy in the identification of key concepts than standard weighting methods such as inverse document frequency. Finally, we propose a probabilistic model for integrating the weighted key concepts identified by our method into a query, and demonstrate that this integration significantly improves retrieval effectiveness for a large set of natural language description queries derived from TREC topics on several newswire and web collections.

Typical pseudo-relevance feedback methods assume the topretrieved documents are relevant and use these pseudo-relevant documents to expand terms. The initial retrieval set can, however, contain a great deal of noise. In this paper, we present a clusterbased resampling method to select better pseudo-relevant documents based on the relevance model. The main idea is to use document clusters to find dominant documents for the initial retrieval set, and to repeatedly feed the documents to emphasize the core topics of a query. Experimental results on large-scale web TREC collections show significant improvements over the relevance model. For justification of the resampling approach, we examine relevance density of feedback documents. A higher relevance density will result in greater retrieval accuracy, ultimately approaching true relevance feedback. The resampling approach shows higher relevance density than the baseline relevance model on all collections, resulting in better retrieval accuracy in pseudo-relevance feedback. This result indicates that the proposed method is effective for pseudo-relevance feedback.

Text reuse occurs in many different types of documents and for many different reasons. One form of reuse, duplicate or near-duplicate documents, has been a focus of researchers because of its importance in Web search. Local text reuse occurs when sentences, facts or passages, rather than whole documents, are reused and modified. Detecting this type of reuse can be the basis of new tools for text analysis. In this paper, we introduce a new approach to detecting local text reuse and compare it to other approaches. This comparison involves a study of the amount and type of reuse that occurs in real documents, including TREC newswire and blog collections.

Retrieval in a question and answer archive involves finding good answers for a user's question. In contrast to typical document retrieval, a retrieval model for this task can exploit question similarity as well as ranking the associated answers. In this paper, we propose a retrieval model that combines a translation-based language model for the question part with a query likelihood approach for the answer part. The proposed model incorporates word-to-word translation probabilities learned through exploiting different sources of information. Experiments show that the proposed translation based language model for the question part outperforms baseline methods significantly. By combining with the query likelihood language model for the answer part, substantial additional effectiveness improvements are obtained.

One aspect in which retrieving named entities is different from retrieving documents is that the items to be retrieved - persons, locations, organizations - are only indirectly described by documents throughout the collection. Much work has been dedicated to finding references to named entities, in particular to the problems of named entity extraction and disambiguation. However, just as important for retrieval performance is how these snippets of text are combined to build named entity representations.

Identifying redundant information in sentences is useful for several applications such as summarization, document provenance, detecting text reuse and novelty detection. The task of identifying redundant information in sentences is defined as follows: Given a query sentence the task is to retrieve sentences from a given collection that express all or some subset of the information present in the query sentence. Sentence retrieval techniques rank sentences based on some measure of their similarity to a query. The effectiveness of such techniques depends on the similarity measure used to rank sentences. An effective retrieval model should be able to handle low word overlap between query and candidate sentences and go beyond just word overlap. Simple language modeling techniques like query likelihood retrieval have outperformed TF-IDF and word overlap based methods for ranking sentences. In this paper, we compare the performance of sentence retrieval using different language modeling techniques for the problem of identifying redundant information.

Query expansion, in the form of pseudo-relevance feedback or relevance feedback, is a common technique used to improve retrieval effectiveness. Most previous approaches have ignored important issues, such as the role of features and the importance of modeling term dependencies. In this paper, we propose a robust query expansion technique based onthe Markov random field model for information retrieval. The technique, called latent concept expansion, provides a mechanism for modeling term dependencies during expansion. Furthermore, the use of arbitrary features within the model provides a powerful framework for going beyond simple term occurrence features that are implicitly used by most other expansion techniques. We evaluate our technique against relevance models, a state-of-the-art language modeling query expansion technique. Our model demonstrates consistent and significant improvements in retrieval effectiveness across several TREC data sets. We also describe how our technique can be used to generate meaningful multi-term concepts for tasks such as query suggestion/reformulation.

Disk access performance is a major bottleneck in traditional information retrieval systems. Compared to system memory, disk bandwidth is poor, and seek times are worse.

We approach the problem of academic literature search by considering an unpublished manuscript as a query to a search system. We use the text of previous literature as well as the citation graph that connects it to find relevant related material. We evaluate our technique with manual and automatic evaluation methods, and find an order of magnitude improvement in mean average precision as compared to a text similarity baseline.

We investigate the difficult problem of matching semi-structured resumes and jobs in a large scale real-world collection. We compare standard approaches to Structured Relevance Models (SRM), an extensionof relevance-based language model for modeling and retrieving semi-structured documents. Preliminary experiments show that the SRM approach achieved promising performance and performed better than typical unstructured relevance models.

Current prediction techniques, which are generally designed for content-based queries and are typically evaluated on relatively homogenous test collections of small sizes, face serious challenges in web search environments where collections are significantly more heterogeneous and different types of retrieval tasks exist. In this paper, we present three techniques to address these challenges. We focus on performance prediction for two types of queries in web search environments: content-based and Named-Page finding. Our evaluation is mainly performed on the GOV2 collection. In addition to evaluating our models for the two types of queries separately, we consider a more challenging and realistic situation that the two types of queries are mixed together without prior information on query types. To assist prediction under the mixed-query situation, a novel query classifier is adopted. Results show that our prediction of web query performance is substantially more accurate than the current state-of-the-art prediction techniques. Consequently, our paper provides a practical approach to performance prediction in real-world web settings.

No abstract available.

The detection of new information in a document stream is an important component of many potential applications. In this work, a new novelty detection approach based on the identification of sentence level information patterns is proposed. First, the information-pattern concept for novelty detection is presented with the emphasis on new information patterns for general topics (queries) that cannot be simply turned into specific questions whose answers are specific named entities (NEs). Then we elaborate a thorough analysis of sentence level information patterns on data from the TREC novelty tracks, including sentence lengths, named entities, sentence level opinion patterns. This analysis provides guidelines in applying those patterns in novelty detection particularly for the general topics. Finally, a unified pattern-based approach is presented to novelty detection for both general and specific topics. The new method for dealing with general topics will be the focus. Experimental results show that the proposed approach significantly improves the performance of novelty detection for general topics as well as the overall performance for all topics from the 2002-2004 TREC novelty tracks.

Several information organization, access, and filtering systems can benefit from different kind of document representations than those used in traditional Information Retrieval (IR). Topic Detection and Tracking (TDT) is an example of such an application. In this paper we demonstrate that named entities serve as better choices of units for document representation over all words. In order to test this hypothesis we study the effect of words-based and entity-based representations on Story Link Detection (SLD) - a core task in TDT research. The experiments on TDT corpora show that entity-based representations give significant improvements for SLD. We also propose a mechanism to expand the set of named entities used for document representation, which enhances the performance in some cases. We then take a step further and analyze the limitations of using only named entities for the document representation. Our studies and experiments indicate that adding additional topical terms can help in addressing such limitations.

In this paper, we introduce the notion of ranking robustness, which refers to a property of a ranked list of documents that indicates how stable the ranking is in the presence of uncertainty in the ranked documents. We propose a statistical measure called the robustness score to quantify this notion. We demonstrate that the robustness score significantly and consistently correlates with query performance in a variety of TREC test collections including the GOV2 collection. We compare the robustness score with the clarity score method which is the state-of-the-art technique for query performance prediction. Our experimental results show that the robustness score performs better than or at least as good as the clarity score. We find that the clarity score is barely correlated with query performance on the GOV2 collection while the correlation between the robustness score and query performance remains significant. We also notice that a combination of the two usually results in more prediction power.

New types of document collections are being developed by various web services. The service providers keep track of non-textual features such as click counts. In this paper, we present a framework to use non-textual features to predict the quality of documents. We also show our quality measure can be successfully incorporated into the language modeling-based retrieval model. We test our approach on a collection of question and answer pairs gathered from a community based question answering service where people ask and answer questions. Experimental results using our quality measure show a significant improvement over our baseline.

No abstract available.

Search algorithms incorporating some form of topic model have a long history in information retrieval. For example, cluster-based retrieval has been studied since the 60s and has recently produced good results in the language model framework. An approach to building topic models based on a formal generative model of documents, Latent Dirichlet Allocation (LDA), is heavily cited in the machine learning literature, but its feasibility and effectiveness in information retrieval is mostly unknown. In this paper, we study how to efficiently use LDA to improve ad-hoc retrieval. We propose an LDA-based document model within the language modeling framework, and evaluate it on several TREC collections. Gibbs sampling is employed to conduct approximate inference in LDA and the computational complexity is analyzed. We show that improvements over retrieval using cluster-based models can be obtained with reasonable efficiency.

There has recently been a significant increase in the number of community-based question and answer services on the Web where people answer other peoples' questions. These services rapidly build up large archives of questions and answers, and these archives are a valuable linguistic resource. One of the major tasks in a question and answer service is to find questions in the archive that a semantically similar to a user's question. This enables high quality answers from the archive to be retrieved and removes the time lag associated with a community-based system. In this paper, we discuss methods for question retrieval that are based on using the similarity between answers in the archive to estimate probabilities for a translation-based retrieval model. We show that with this model it is possible to find semantically similar questions with relatively little word overlap.

The detection of new information in a document stream is an important component of many potential applications. In this paper, a new novelty detection approach based on the identification of sentence level patterns is proposed. Given a user's information need, some patterns in sentences such as combinations of query words, named entities and phrases, may contain more important and relevant information than single words. Therefore, the proposed novelty detection approach focuses on the identification of previously unseen query-related patterns in sentences. Specifically, a query is preprocessed and represented with patterns that include both query words and required answer types. These patterns are used to retrieve sentences, which are then determined to be novel if it is likely that a new answer is present. An analysis of patterns in sentences was performed with data from the TREC 2002 novelty track and experiments on novelty detection were carried out on data from the TREC 2003 and 2004 novelty tracks. The experimental results show that the proposed pattern-based approach significantly outperforms all three baselines in terms of precision at top ranks.

No abstract available.

Text similarity spans a spectrum, with broad topical similarity near one extreme and document identity at the other. Intermediate levels of similarity -- resulting from summarization, paraphrasing, copying, and stronger forms of topical relevance -- are useful for applications such as information flow analysis and question-answering tasks. In this paper, we explore mechanisms for measuring such intermediate kinds of similarity, focusing on the task of identifying where a particular piece of information originated. We consider both sentence-to-sentence and document-to-document comparison, and have incorporated these algorithms into RECAP, a prototype information flow analysis tool. Our experimental results with RECAP indicate that new mechanisms such as those we propose are likely to be more appropriate than existing methods for identifying the intermediate forms of similarity.

The quality of document content, which is an issue that is usually ignored for the traditional ad hoc retrieval task, is a critical issue for Web search. Web pages have a huge variation in quality relative to, for example, newswire articles. To address this problem, we propose a document quality language model approach that is incorporated into the basic query likelihood retrieval model in the form of a prior probability. Our results demonstrate that, on average, the new model is significantly better than the baseline (query likelihood model) in terms of precision at the top ranks.

A large number of question and answer pairs can be collected from question and answer boards and FAQ pages on the Web. This paper proposes an automatic method of finding the questions that have the same meaning. The method can detect semantically similar questions that have little word overlap because it calculates question-question similarities by using the corresponding answers as well as the questions. We develop two different similarity measures based on language modeling and compare them with the traditional similarity measures. Experimental results show that semantically similar questions pairs can be effectively found with the proposed similarity measures.

No abstract available.

This paper develops a general, formal framework for modeling term dependencies via Markov random fields. The model allows for arbitrary text features to be incorporated as evidence. In particular, we make use of features based on occurrences of single terms, ordered phrases, and unordered phrases. We explore full independence, sequential dependence, and full dependence variants of the model. A novel approach is developed to train the model that directly maximizes the mean average precision rather than maximizing the likelihood of the training data. Ad hoc retrieval experiments are presented on several newswire and web collections, including the GOV2 collection used at the TREC 2004 Terabyte Track. The results show significant improvements are possible by modeling dependencies, especially on the larger web collections.

Previous research into the efficiency of text retrieval systems has dealt primarily with methods that consider inverted lists in sequence; these methods are known as term-at-a-time methods. However, the literature for optimizing document-at-a-time systems remains sparse.We present an improvement to the max_score optimization, which is the most efficient known document-at-a-time scoring method. Like max_score, our technique, called term bounded max_score, is guaranteed to return exactly the same scores and documents as an unoptimized evaluation, which is particularly useful for query model research. We simulated our technique to explore the problem space, then implemented it in Indri, our large scale language modeling search engine. Tests with the GOV2 corpus on title queries show our method to be 23% faster than max_score alone, and 61% faster than our document-at-a-time baseline. Our optimized query times are competitive with conventional term-at-a-time systems on this year's TREC Terabyte task.

Query expansion is a well-known technique that has been shown to improve average retrieval performance. This technique has not been used in many operational systems because of the fact that it can greatly degrade the performance of some individual queries. We show how comparison between language models of the unexpanded and expanded retrieval results can be used to predict when the expanded retrieval has strayed from the original sense of the query. In these cases, the unexpanded results are used while the expanded results are used in the remaining cases (where such straying is not detected). We evaluate this method on a wide variety of TREC collections.

Answer patterns have been shown to improve the perfor-mance of open-domain factoid QA systems. Their use, however, requires either constructing the patterns manually or developing algorithms for learning them automatically. We present here a simpler approach that extends the techniques of language modeling to create answer models. These are language models trained on the correct answers to training questions. We show how they fit naturally into a probabilis-tic model for answer passage retrieval and demonstrate their effectiveness on the TREC 2002 QA Corpus.

Previous research on cluster-based retrieval has been inconclusive as to whether it does bring improved retrieval effectiveness over document-based retrieval. Recent developments in the language modeling approach to IR have motivated us to re-examine this problem within this new retrieval framework. We propose two new models for cluster-based retrieval and evaluate them on several TREC collections. We show that cluster-based retrieval can perform consistently across collections of realistic size, and significant improvements over document-based retrieval can be obtained in a fully automatic manner and without relevance information provided by human.

No abstract available.

No abstract available.

ABSTRACT Although information retrieval research has always been concernedwith improving the effectiveness of search, in some applications, such as information analysis, a more specific requirement exists forhigh accuracy retrieval. This means that achieving high precision in the top document ranks is paramount. In this paper we presentwork aimed at achieving high accuracy in ad-hoc document retrieval by incorporating approaches from question answering (QA).We focus on getting the first relevant result as high as possible in the ranked list and argue that traditional precision and recall are notappropriate measures for evaluating this task. We instead use the mean reciprocal rank (MRR) of the first relevant result. We eval-uate three different methods for modifying queries to achieve high accuracy. The experiments done on TREC data provide support forthe approach of using MRR and incorporating QA techniques for getting high accuracy in ad-hoc retrieval task. Categories and Subject Descriptors H.3.4 [Information Storage and Retrieval]: Systems and Soft-ware--Performance evaluation (efficiency and effectiveness) ; H.3.3 [Information Storage and Retrieval]: Information Search andRetrieval--Query formulation

We explore the relationship between time and relevance using TREC ad-hoc queries. A type of query is identified that favors very recent documents. We propose a time-based language model approach to retrieval for these queries. We show how time can be incorporated into both query-likelihood models and relevance models. These models were used for experiments comparing time-based language models to heuristic techniques for incorporating document recency in the ranking. Our results show that time-based models perform as well as or better than the best of the heuristic techniques.

In traditional relevance feedback, researchers have explored relevant document feedback, wherein, the query representation is updated based on a set of relevant documents returned by the user. In this work, we investigate relevant query feedback, in which we update a document's representation based on a set of relevant queries. We propose four statistical models to incorporate relevant query feedback.To validate our models, we considered anchor text of incoming links to a given document as feedback queries and performed experiments on the home-page retrieval task of TREC 2001. Our results show that three of our four models outperform the query-likelihood baseline by at least 35% in MRR score on a test set.

One of the major challenges in the field of information retrieval (IR) is to specify a formal framework that both describes the important processes involved in finding relevant information, and successfully predicts which techniques will provide good effectiveness in terms of accuracy. A recent approach that has shown considerable promise uses generative models of text (language models) to describe the IR processes. We briefly review the major variations of the language model approach and how they have been used to develop a range of retrieval-related language technologies, including cross-lingual IR and distributed search. We also discuss how this approach could be used with structured data extracted from text.

Following the tradition of these acceptance talks, I will be giving my thoughts on where our field is going. Any discussion of the future of information retrieval (IR) research, however, needs to be placed in the context of its history and relationship to other fields. Although IR has had a very strong relationship with library and information science, its relationship to computer science (CS) and its relative standing as a sub-discipline of CS has been more dynamic. IR is quite an old field, and when a number of CS departments were forming in the 60s, it was not uncommon for a faculty member to be pursuing research related to IR. Early ACM curriculum recommendations for CS contained courses on information retrieval, and encyclopedias described IR and database systems as different aspects of the same field.

Hierarchies provide a means of organizing, summarizing and accessing information. We describe a method for automatically generating hierarchies from small collections of text, and then apply this technique to summarizing the documents retrieved by a search engine.

The ability to find tables and extract information from them is a necessary component of data mining, question answering, and other information retrieval tasks. Documents often contain tables in order to communicate densely packed, multi-dimensional information. Tables do this by employing layout patterns to efficiently indicate fields and records in two-dimensional form.Their rich combination of formatting and content present difficulties for traditional language modeling techniques, however. This paper presents the use of conditional random fields (CRFs) for table extraction, and compares them with hidden Markov models (HMMs). Unlike HMMs, CRFs support the use of many rich and overlapping layout and language features, and as a result, they perform significantly better. We show experimental results on plain-text government statistical reports in which tables are located with 92% F1, and their constituent lines are classified into 12 table-related categories with 94% accuracy. We also discuss future work on undirected graphical models for segmenting columns, finding cells, and classifying them as data cells or label cells.

We describe results from an ongoing project that considers question types and document features and their relationship to retrieval techniques. We examine eight document features from the top 25 documents retrieved from 74 questions and find that lists and FAQs occur in more documents judged relevant to task-oriented questions than those judged relevant to fact-oriented questions.

Previous research has shown that passage-level evidence can bring added benefits to document retrieval when documents are long or span different subject areas. Recent developments in language modeling approach to IR provided a new effective alternative to traditional retrieval models. These two streams of research motivate us to examine the use of passages in a language model framework. This paper reports on experiments using passages in a simple language model and a relevance model, and compares the results with document-based retrieval. Results from the INQUERY search engine, which is not based on a language modeling approach, are also given for comparison. Test data include two heterogeneous and one homogeneous document collections. Our experiments show that passage retrieval is feasible in the language modeling context, and more importantly, it can provide more reliable performance than retrieval based on full documents.

We develop a method for predicting query performance by computing the relative entropy between a query language model and the corresponding collection language model. The resulting clarity score measures the coherence of the language usage in documents whose models are likely to generate the query. We suggest that clarity scores measure the ambiguity of a query with respect to a collection of documents and show that they correlate positively with average precision in a variety of TREC test sets. Thus, the clarity score may be used to identify ine#ective queries, on average, without relevance information. We develop an algorithm for automatically setting the clarity score threshold between predicted poorly-performing queries and acceptable queries and validate it using TREC data. In particular, we compare the automatic thresholds to optimum thresholds and also check how frequently results as good are achieved in sampling experiments that randomly assign queries to the two classes.

We propose a formal model of Cross-Language Information Retrieval that does not rely on either query translation or document translation. Our approach leverages recent advances in language modeling to directly estimate an accurate topic model in the target language, starting with a query in the source language. The model integrates popular techniques of disambiguation and query expansion in a unified formal framework. We describe how the topic model can be estimated with either a parallel corpus or a dictionary. We test the framework by constructing Chinese topic models from English queries and using them in the CLIR task of TREC9. The model achieves performance around 95% of the strong mono-lingual baseline in terms of average precision. In initial precision, our model outperforms the mono-lingual baseline by 20%. The main contribution of this work is the unified formal model which integrates techniques that are essential for effective Cross-Language Retrieval.

No abstract available.

Abstract XML holds the promise to yield (1) a more precise search by providing additional information in the elements, (2) a better integrated search of documents from heterogeneous sources, (3) a powerful search paradigm using structural as well as content specifications, and (4) data and information exchange to share resources and to support cooperative search. We survey several indexing techniques for XML documents, grouping them into flat‐file, semistructured, and structured indexing paradigms. Searching techniques and supporting techniques for searching are reviewed, including full text search and multistage search. Because searching XML documents can be very flexible, various search result presentations are discussed, as well as database and information retrieval system integration and XML query languages. We also survey various retrieval models, examining how they would be used or extended for retrieving XML documents. To conclude the article, we discuss various open issues that XML poses with respect to information retrieval and database research.

We explore the relation between classical probabilistic models of information retrieval and the emerging language modeling approaches. It has long been recognized that the primary obstacle to effective performance of classical models is the need to estimate arelevance model: probabilities of words in the relevant class. We propose a novel technique for estimating these probabilities using the query alone. We demonstrate that our technique can produce highly accurate relevance models, addressing important notions of synonymy and polysemy. Our experiments show relevance models outperforming baseline language modeling systems on TREC retrieval and TDT tracking tasks. The main contribution of this work is an effective formal method for estimating a relevance model with no training data.

Hierarchies have long been used for organization, summarization, and access to information. In this paper we define summarization in terms of a probabilistic language model and use the definition to explore a new technique for automatically generating topic hierarchies by applying a graph-theoretic algorithm, which is an approximation of the Dominating Set Problem. The algorithm efficiently chooses terms according to a language model. We compare the new technique to previous methods proposed for constructing topic hierarchies including subsumption and lexical hierarchies, as well as the top TF.IDF terms. Our results show that the new technique consistently performs as well as or better than these other techniques. They also show the usefulness of hierarchies compared with a list of terms.

Techniques for automatic query expansion have been extensively studied in information research as a means of addressing the word mismatch between queries and documents. These techniques can be categorized as either global or local. While global techniques rely on analysis of a whole collection to discover word relationships, local techniques emphasize analysis of the top-ranked documents retrieved for a query. While local techniques have shown to be more effective that global techniques in general, existing local techniques are not robust and can seriously hurt retrieved when few of the retrieval documents are relevant. We propose a new technique, called local context analysis, which selects expansion terms based on cooccurrence with the query terms within the top-ranked documents. Experiments on a number of collections, both English and non-English, show that local context analysis offers more effective and consistent retrieval results.

Statistical language modeling has been successfully used for speech recognition, part-of-speech tagging, and syntactic parsing. Recently, it has also been applied to information retrieval. According to this new paradigm, each document is viewed as a language sample, and a query as a generation process. The retrieved documents are ranked based on the probabilities of producing a query from the corresponding language models of these documents. In this paper, we will present a new language model for information retrieval, which is based on a range of data smoothing techniques, including the Good-Turning estimate, curve-fitting functions, and model combinations. Our model is conceptually simple and intuitive, and can be easily extended to incorporate probabilities of phrases such as word pairs and word triples. The experiments with the Wall Street Journal and TREC4 data sets showed that the performance of our model is comparable to that of INQUERY and better than that of another language model for information retrieval. In particular, word pairs are shown to be useful in improving the retrieval performance.

This paper presents a means of automatically deriving a hierarchical organization of concepts from a set of documents without use of training data or standard clustering techniques.Instead, salient words and phrases extracted from the documents are organized hierarchically using a type of co-occurrence known as subsumption.The resulting structure is displayed as a series of hierarchical menus.When generated from a set of retrieved documents, a user browsing the menus is provided with a detailed overview of their content in a manner distinct from existing overview and summarization techniques.The methods used to build the structure are simple, but appear to be effective: a smallscale user study reveals that the generated hierarchy possesses properties expected of such a structure in that general terms are placed at the top levels leading to related and more specific terms below.The formation and presentation of the hierarchy is described along with the user study and some other informal evaluations.

No abstract available.

Effective retrieval in a distributed environment is an important but difficult problem. Lack of effectiveness appears to have three causes. First, collection selection based on word histograms is not appropriate for heterogeneous collections. Second, relevant documents are scattered over many collections and searching a few collections misses many relevant documents. Third, most existing collection selection metrics lack sound theoretical justifications and hence may not be well tuned to the problem. We propose a new approach to distributed retrieval based on document clustering and language modeling. Document clustering is used to organize collections around topics. Language modeling is used to properly represent topics and effectively select the right topics for a query. Based on these ideas, three methods are proposed to suit different environments. We show that all three methods improve effectiveness of distributed retrieval. 1 Introduction Information has become highly distribut...

The inference network model of information retrieval allows a probabilistic interpretation of query operators. In particular, Boolean query operators are conveniently modeled as link matrices of the Bayesian Network. Prior work has shown, however, that these operators do not perform as well as the pnorm operators used for modeling query operators in the context of the vector space model. This motivates the search for alternative probabilistic formulations for these operators. The design of such alternatives must contend with the issue of computational tractability, since the evaluation of an arbitrary operator requires exponential time. We define a flexible class of link matrices that are natural candidates for the implementation of query operators and an O ( n 2 ) algorithm ( n = the number of parent nodes) for the computation of probabilities involving link matrices of this class. We present experimental results indicating that Boolean operators implemented in terms of link matrices from this class perform as well as pnorm operators in the context of the INQUERY inference network.

One of the main hurdles to improved CLIR effectiveness is resolving ambiguity associated with translation.Availability of resources is also a problem.First we present a technique based on co-occurrence statistics from unlinked corpora which can be used to reduce the ambiguity associated with phrasal and term translation.We then combine this method with other techniques for reducing ambiguity and achieve more than 90% monolingual effectiveness.Finally, we compare the co-occurrence method with parallel corpus and machine translation techniques and show that good retrieval effectiveness can be achieved without complex resources.

Article A language modeling approach to information retrieval Share on Authors: Jay M. Ponte Computer Science Department, University of Massachusetts, Amherst Computer Science Department, University of Massachusetts, AmherstView Profile , W. Bruce Croft Computer Science Department, University of Massachusetts, Amherst Computer Science Department, University of Massachusetts, AmherstView Profile Authors Info & Claims SIGIR '98: Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrievalAugust 1998 Pages 275–281https://doi.org/10.1145/290941.291008Online:01 August 1998Publication History 1,592citation7,260DownloadsMetricsTotal Citations1,592Total Downloads7,260Last 12 Months258Last 6 weeks34 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Stemming is used in many information retrieval (IR) systems to reduce variant word forms to common roots. It is one of the simplest applications of natural-language processing to IR and is one of the most effective in terms of user acceptance and consistency, though small retrieval improvements. Current stemming techniques do not, however, reflect the language use in specific corpora, and this can lead to occasional serious retrieval failures. We propose a technique for using corpus-based word variant cooccurrence statistics to modify or create a stemmer. The experimental results generated using English newspaper and legal text and Spanish text demonstrate the viability of this technique and its advantages relative to conventional approaches that only employ morphological rules.

Dictionary methods for cross-language information retrieval give performance below that for mono-lingual retrieval.Failure to translate multi-term phrases has km shown to be one of the factors responsible for the errors associated with dictionary methods.First, we study the importance of phrasaI translation for this approach.Second, we explore the role of phrases in query expansion via local context analysis and local feedback and show how they can be used to significantly reduce the error associated with automatic dictionary translation.

article Free Access Share on Computationally tractable probabilistic modeling of Boolean operators Authors: Warren R. Greiff Computer Science Department, University of Masaachusetts, Amherst, Amherst, Massachusetts Computer Science Department, University of Masaachusetts, Amherst, Amherst, MassachusettsView Profile , W. Bruce Croft Computer Science Department, University of Masaachusetts, Amherst, Amherst, Massachusetts Computer Science Department, University of Masaachusetts, Amherst, Amherst, MassachusettsView Profile , Howard Turtle West Publishing Co., Eagan, Minnesota West Publishing Co., Eagan, MinnesotaView Profile Authors Info & Claims ACM SIGIR ForumVolume 31Issue SIDecember 1997 pp 119–128https://doi.org/10.1145/278459.258547Published:01 July 1997Publication History 4citation480DownloadsMetricsTotal Citations4Total Downloads480Last 12 Months23Last 6 weeks5 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my Alerts New Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

Three different types of classifiers were investigated in the context of a text categorization problem in the medical domain: the automatic assignment of ICD9 codes to dictated inpatient discharge summaries. K-nearest-neighbor, relevance feedback, and Bayesian independence classifers were applied individually and in combination. A combination of different classifiers produced better results than any single type of classifier. For this specific medical categorization problem, new query formulation and weighting methods used in the k-nearest-neighbor classifier improved performance. 1 Introduction Past research in information retrieval has shown that one can improve retrieval effectiveness by using multiple representations in indexing and query formulation [27] [19] [3] [11] and by using multiple search strategies [5] [24] [7]. In this work, we investigate whether we can attain similar improvements in the domain of text categorization by combining different representations and classif...

Automatic query expansion has long been suggested as a technique for dealing with the fundamental issue of word mismatch in information retrieval.A number of approaches to ezpanrnion have been studied and, more recently, attention has focused on techniques that analyze the corpus to discover word relationship (global techniques) and those that analyze documents retrieved by the initial quer~( local feedback).In this paper, we compare the effectiveness of these approaches and show that, although global analysis haa some advantages, local analysia is generally more effective.We also show that using global analysis techniques, such as word contezt and phrase structure, on the local aet of documents produces results that are both more effective and more predictable than simple local feedback.

The use of information retrieval systems in networked environments raises a new set of issues that have received little attention.These issues include ranking document collections for relevance to a query, selecting the best set of collections from a ranked list, and merging the document rankings that are returned from a set of collections.This paper describes methods of addressing each issue in the inference network model, dkcusses their implementation in the INQUERY system, and presents experimental results demonstrating their effectiveness.1

The full integration of information retrieval (IR) features into a database management system (DBMS) has long been recognized as both a significant goal and a challenging undertaking. By full integration we mean: i) support for document storage, indexing, retrieval, and update, ii) transaction semantics, thus all database operations on documents have the ACID properties of atomicity, consistency, isolation, and durability, iii) concurrent addition, update, and retrieval of documents, and iv) database query language extensions to provide ranking for document retrieval operations. It is also necessary for the integrated offering to exhibit scaleable performance for document indexing and retrieval processes. To identify the implementation requirements imposed by the desired level of integration, we layered a representative IR application on Oracle Rdb and then conducted a number of database load and document retrieval experiments. The results of these experiments suggest that infrastructural extensions are necessary to obtain both the desired level of IR integration and scaleable performance. With the insight gained from our initial experiments, we developed an approach, called cooperative indexing, that provides a framework to achieve both scaleability and full integration of IR and RDBMS technology. Prototype implementations of system-level extensions to support cooperative indexing were evaluated with a modified version of Oracle Rdb. Our experimental findings validate the cooperative indexing scheme and suggest alternatives to further improve performance.

Results from research in information retrieval have suggested that significant improvements in retrieval effectiveness can be obtained by combining results from multiple index representations, query formulations, and search strategies. The inference net model of retrieval, which was designed from this point of view, treats information retrieval as an evidential reasoning process where multiple sources of evidence about document and query content are combined to estimate relevance probabilities. In this article, we use a system based on this model to study the retrieval effectiveness benefits of combining the types of document and query information that are found in typical commercial databases and information services. The results indicate that substantial real benefits are possible. © 1995 John Wiley & Sons, Inc.

Five independently generated Boolean query formulations for ten different TREC topics were produced by ten different expert online searchers. These different formulations were grouped, and the groups, and combinations of them, were used as searches against the TREC test collection, using the INQUERY probabilistic inference network retrieval engine. Results show that progressive combination of query formulations leads to progressively improving retrieval performance. Results were compared against the performance of INQUERY natural language based queries, and in combination with them. The issue of recall as a performance measure in large databases was raised, since overlap between the searches conducted in this study, and the TREC-1 searches, was smaller than expected.

The TIPSTER collection is unusual because of both its size and detail. In particular, it describes a set of information needs, as opposed to traditional queries. These detailed representations of information need are an opportunity for research on different methods of formulating queries. This paper describes several methods of constructing queries for the INQUERY information retrieval system, and then evaluates those methods on the TIPSTER document collection. Both AdHoc and Routing query processing methods are evaluated.

A series of Japanese full-text retrieval experiments were conducted using an inference network document retrieval model. The retrieval performance of two major indexing methods, character-based and word-based, were evaluated. Using structured queries, the character-based indexing performed retrieval as well as, or slightly better, than the word-based system. This result has practical significance since the character-based indexing speed is considerably faster than the traditional word-based indexing. All the queries in this experiment were automatically formulated from natural language input.

Relevance feedback, which modifies queries using judgements of the relevance of a few, highly-ranked documents, has historically been an important method for increasing the performance of information retrieval systems. In this paper, we extend the inference network model introduced by Turtle and Croft to include relevance feedback techniques. The difference between relevance feedback on text abstracts and full text collections is studied. Preliminary results for relevance feedback on the structured queries supported by the inference net model are also reported.

Document management systems are needed for many business applications. This type of system would combine the functionality of a database system, (for describing, storing and maintaining documents with complex structure and relationships) with a text retrieval system (for effective retrieval based on full text). The retrieval model for a document management system is complicated by the variety and complexity of the objects that are represented. In this paper, we describe an approach to complex object retrieval using a probabilistic inference net model, and an implementation of this approach using a loose coupling of an object-oriented database system (IRIS) and a text retrieval system based on inference nets (INQUERY). The resulting system is used to store long, structured documents and can retrieve document components (sections, figures, etc.) based on their contents or the contents of related components. The lessons learnt from the implementation are discussed.

Lexical ambiguity is a pervasive problem in natural language processing. However, little quantitative information is available about the extent of the problem or about the impact that it has on information retrieval systems. We report on an analysis of lexical ambiguity in information retrieval test collections and on experiments to determine the utility of word meanings for separating relevant from nonrelevant documents. The experiments show that there is considerable ambiguity even in a specialized database. Word senses provide a significant separation between relevant and nonrelevant documents, but several factors contribute to determining whether disambiguation will make an improvement in performance. For example, resolving lexical ambiguity was found to have little impact on retrieval effectiveness for documents that have many words in common with the query. Other uses of word sense disambiguation in an information retrieval context are discussed.

Both phrases and Boolean queries have a long history in information retrieval, particularly in commercial systems.In previous work, Boolean queries have been used as a source of phrases for a statistical retrieval model, This work, like the majority of research on phrases, resulted in little improvement in retrieval effectiveness, In this paper, we describe an approach where phrases identified in natural language queries are used to build structured queries for a probabilistic retrieval model.Our results show that using phrases in this way can improve performance, and that phrases that are automatically extracted from a natural language query perform nearly as well as manually selected phrases.

In some recent experimental document retrieval systems, emphasis has been placed on the acquisition of a detailed model of the information need through interaction with the user. It has been argued that these “enhanced” queries, in combination with relevance feedback, will improve retrieval performance. In this paper, we describe a study with the aim of evaluating how easily enhanced queries can be acquired from users and how effectively this additional knowledge can be used in retrieval. The results indicate that significant effectiveness benefits can be obtained through the acquisition of domain concepts related to query concepts, together with their level of importance to the information need.

Term clustering and syntactic phrase formation are methods for transforming natural language text. Both have had only mixed success as strategies for improving the quality of text representations for document retrieval. Since the strengths of these methods are complementary, we have explored combining them to produce superior representations. In this paper we discuss our implementation of a syntactic phrase generator, as well as our preliminary experiments with producing phrase clusters. These experiments show small improvements in retrieval effectiveness resulting from the use of phrase clusters, but it is clear that corpora much larger than standard information retrieval test collections will be required to thoroughly evaluate the use of this technique.

The use of inference networks to support document retrieval is introduced. A network-based retrieval model is described and compared to conventional probabilistic and Boolean models.

Most approachesto full-text information retrieval currently index documents based on the words they contain, and retrieve them based on the word's frequency of occurrence.This can cause many irrelevant documents to be retrieved because words are often ambiguous.We propose an approach in which documents are indexed by word aenaea, and in which these senses are taken from a machine-readable dictionary.We review some of the work on machine-readable dictionaries and the approaches that have been taken to word sense disambiguation.We then discuss our own approach to the problem based on the use of multiple sources of evidence.We conclude with the results of some experiments that indicate the degree to which lexical ambiguity is a factor in current systems.

There is no task that computers regularly perform that is more affected by the nature of human language than the retrieval of texts in response to a human need. Despite this, the techniques actually in use for this task, as well as most of the techniques proposed by information retrieval (IR) researchers, make little use of knowledge about language. In this article we take the view that IR is an inference task, and that natural language processing (NLP) techniques can produce text representations that enable more accurate inferences about document content. By considering previous work on language-based and knowledge-based techniques from this perspective, some clear lessons are apparent, and we are applying these lessons in the ADRENAL (Augmented Document REtrieval using NAtural Language processing) project. Our initial experiments with hand-coded representations suggest that using NLP-produced representations can result in significant performance increases in IR systems, and also demonstrate the attention that must be given to representational issues in language-oriented IR.

Choosing an appropriate document representation and search strategy for document retrieval has been largely guided by achieving good average performance instead of optimizing the results for each individual query. A model of retrieval based on plausible inference gives us a different perspective and suggests that techniques should be found for combining multiple sources of evidence (or search strategies) into an overall assessment of a document's relevance, rather than attempting to pick a single strategy. In this paper, we explain our approach to plausible inference for retrieval and describe some preliminary experiments designed to test this approach. The experiments use a spreading activation search to implement the plausible inference process. The results show that significant effectiveness improvements are possible using this approach.

Signature files provide an efficient access method for text in documents, but retrieval is usually limited to finding documents that contain a specified Boolean pattern of words. Effective retrieval requires that documents with similar meanings be found through a process of plausible inference. The simplest way of implementing this retrieval process is to rank documents in order of their probability of relevance. In this paper techniques are described for implementing probabilistic ranking strategies with sequential and bit-sliced signature tiles and the limitations of these implementations with regard to their effectiveness are pointed out. A detailed comparison is made between signature-based ranking techniques and ranking using term-based document representatives and inverted files. The comparison shows that term-based representations are at least competitive (in terms of efficiency) with signature files and, in some situations, superior.

Document retrieval systems have been restricted, by the nature of the task, to techniques that can be used with large numbers of documents and broad domains. The most effective techniques that have been developed are based on the statistics of word occurrences in text. In this paper, we describe an approach to using natural language processing (NLP) techniques for what is essentially a natural language problem - the comparison of a request text with the text of document titles and abstracts. The proposed NLP techniques are used to develop a request model based on “conceptual case frames” and to compare this model with the texts of candidate documents. The request model is also used to provide information to statistical search techniques that identify the candidate documents. As part of a preliminary evaluation of this approach, case frame representations of a set of requests from the CACM collection were constructed. Statistical searches carried out using dependency and relative importance information derived from the request models indicate that performance benefits can be obtained.

Data models have been proposed as a means of defining the objects and operations in an office information system. Office documents, because of their highly variable structure and multimedia content, are a difficult class of objects to model. The modeling task is further complicated by document architecture standards used for interchange between systems. We present an approach to data modeling based on constrained type definitions that allows architecture standards to be defined and ensures that individual document types conform to those standards. The ADABTPL model, which is used to define the schema of document types and standards, is described.

The most effective method of improving the retrieval performance of a document retrieval system is to acquire a detailed specification of the user's information need. The system described in this article, I3R, provides a number of facilities and search strategies based on this approach. The system uses a novel architecture to allow more than one system facility to be used at a given stage of a search session. Users influence the system actions by stating goals they wish to achieve, by evaluating system output, and by choosing particular facilities directly. The other main features of I3R are an emphasis on domain knowledge used for refining the model of the information need, and the provision of a browsing mechanism that allows the user to navigate through the knowledge base. © 1987 John Wiley & Sons, Inc.

The introduction of domain knowledge into a document retrieval system has two important consequences; an increase in the effectiveness of retrieval and a decrease in the efficiency of text processing. In this paper, a method is presented of combining user-specified domain knowledge with efficient retrieval techniques based on probabilistic models. The domain knowledge is represented as a collection of frames that contain rules specifying recognition conditions for domain concepts and relationships between concepts. The inference network represented in these frames is used to infer the concepts that are related to a user's query. This approach is being implemented as part of the I3R expert intermediary system.

A method of integrating Boolean queries with probabilistic retrieval models is proposed. Boolean queries are interpreted as specifying term dependencies that can be used to correct the document scores obtained with a basic probabilistic model. Alternative methods of obtaining dependency information, such as user-specified phrases, can also be used in this approach. The experimental results indicate that significant performance benefits can be obtained, particularly when dependencies are derived from term phrases identified in natural language queries. © 1986 John Wiley & Sons, Inc.

A major goal of an office system is to support tasks that are central to office functions.Some office tasks are readily implemented with generic office tools, such as calendars, forms packages, and mail.Many tasks, however, involve complex sequences of actions which do not all correspond to tool invocations but, instead, rely on the problem-solving abilities of office workers.In this paper we describe a system (POISE) than can be used to both automate routine tasks and provide assistance in more complex situations.The type of assistance provided can range from maintaining a record of the tasks currently being executed to suggesting possible next steps and answering natural language queries about the tasks.The POISE system uses both a procedure-based and a goal-based representation of the tasks to achieve efficiency and flexibility.The mechanisms used by POISE are described with example procedures from a university office.

Article Free Access Share on Applications for information retrieval techniques in the office Author: W. Bruce Croft University of Massachusetts, Amherst, MA University of Massachusetts, Amherst, MAView Profile Authors Info & Claims SIGIR '83: Proceedings of the 6th annual international ACM SIGIR conference on Research and development in information retrievalJune 1983 Pages 18–23https://doi.org/10.1145/511793.511797Published:01 June 1983Publication History 5citation390DownloadsMetricsTotal Citations5Total Downloads390Last 12 Months89Last 6 weeks17 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

A network organization for implementing a document retrieval system is proposed. This organization has significant advantages in terms of the range of searches that can be used when compared to either inverted or clustered file organizations. Algorithms for generating and maintaining the network are described together with experiments designed to test their efficiency and effectiveness.

Many effective search strategies derived from different models are available for document retrieval systems. However, it does not appear that there is a single most effective strategy. Instead, different strategies perform optimally under different conditions. This paper outlines the design of an adaptive document retrieval system that chooses the best search strategy for a particular situation and user. In order to be able to support a variety of search strategies, a general network representation of the documents and terms in the database is proposed. This network representation leads to efficient methods of generating and using document and term classifications.One of the most desirable features of an adaptive system would be the ability to learn from experience. A method of incorporating this learning ability into the system is described. The adaptive control strategy for choosing search strategies enables the system to base its actions on a number of factors, including a model of the current user.Finally, some ideas for a flexible interface for casual users are suggested. Part of this interface is the heuristic search, which is used when searches based on formal models have failed. The heuristic search provides a browsing capability for the user.

Abstract Probabilistic models of retrieval have provided insights into the document retrieval process and contain the basis for very effective search strategies. A major limitation of these models is that they assume that documents are represented by binary index terms. In many cases the index terms will be assigned weights, such as within‐document frequency weights, which are derived from the content of the documents by the indexing process. These weights, which are referred to here as term significance weights, indicate the relative importance of the terms in individual documents. This article describes how retrieval models which use either independence or dependence assumptions can be extended to include document representatives containing term significance weights. Comparison with other research indicates that search strategies based on models modified in this way can further improve the effectiveness of document retrieval systems.

Recently several models of the search process in a document retrieval system have been proposed and retrieval experiments have shown that they will improve system performance. These include models which use relevance judgements to rank documents in order of probability of relevance and models of retrieval from clusters of documents. In this paper various models are compared in terms of the ease with which they could be implemented. An important consideration is how this implementation would be affected by the introduction of new hardware such as content-addressable memories. The main conclusion is that models which concentrate on improving the effectiveness of the search process are not rendered redundant by the availability of new hardware. However, the efficiency of their implementation would be improved.

Abstract Various ways of using Boolean queries with a clustered file organization are described. An experiment indicates that such retrieval methods can be effective.

A file organization for cluster-based retrieval is presented and tested. This file organization is based on the bottom-up search which, in contrast to the more usual top-down search, starts at the lowest level of a cluster hierarchy (the documents) and looks at progressively larger clusters. This approach enables most of the efficiency problems previously associated with clustered file organizations to be avoided. There are two parts to this file organization - a compact cluster hierarchy representation which does not store cluster representatives and a compact inverted file which is used to provide a starting point for the bottom-up search.

Abstract A method for clustering large files of documents using a clustering algorithm which takes O(n2) operations (single‐link) is proposed. This method is tested on a file of 11,613 documents derived from an operational system. One property of the generated cluster hierarchy (hierarchy connection percentage) is examined and it indicates that the hierarchy is similar to those from other test collections. A comparison of clustering times with other methods showsthat large files can be clustered by single‐link in a time at least comparable to various heuristic algorithms which theoretically require fewer operations.