Exploratory searches are characterized by under-specified goals and evolving query intents. In such scenarios, retrieval models that can capture user-specified nuances in query intent and adapt results accordingly are desirable — instruction-following retrieval models promise such a capability. In this work, we evaluate instructed retrievers for the prevalent yet under-explored application of aspect-conditional seed-guided exploration using an expert-annotated test collection. We evaluate both recent LLMs fine-tuned for instructed retrieval and general-purpose LLMs prompted for ranking with the highly performant Pairwise Ranking Prompting. We find that the best instructed retrievers improve on ranking relevance compared to instruction-agnostic approaches. However, we also find that instruction following performance, crucial to the user experience of interacting with models, does not mirror ranking relevance improvements and displays insensitivity or counter-intuitive behavior to instructions. Our results indicate that while users may benefit from using current instructed retrievers over instruction-agnostic models, they may not benefit from using them for long-running exploratory sessions requiring greater sensitivity to instructions.

Personalization in Question Answering (QA) requires answers that are both accurate and aligned with users' background, preferences, and historical context. Existing state-of-the-art methods primarily rely on retrieval-augmented generation (RAG) solutions that construct personal context by retrieving relevant items from the user's profile. Existing methods use the user's query directly to retrieve personal documents and such strategies often lead to surface-level personalization. We propose PR2 (Personalized Retrieval-Augmented Reasoning), a reinforcement learning framework that integrates reasoning and retrieval from personal context for personalization. PR2 learns adaptive retrieval-reasoning policies, determining when to retrieve, what evidence to retrieve from user profiles, and how to incorporate it into intermediate reasoning steps. By optimizing multi-turn reasoning trajectories under a personalized reward function, the framework reinforces reasoning paths that better align with user-specific preferences and contextual signals reflected by the reward model. Extensive experiments on the LaMP-QA benchmark using three LLMs show that PR2 consistently outperforms strong baselines, achieving an average relative improvement of 8.26%-12.0% in personalized QA.

Dense retrieval, which encodes queries and documents into a single dense vector, has become the dominant neural retrieval approach due to its simplicity and compatibility with fast approximate nearest neighbor algorithms. As the tasks dense retrieval performs grow in complexity, the fundamental limitations of the underlying data structure and similarity metric---namely vectors and inner-products---become more apparent. Recent work has shown theoretical limitations inherent to single vectors and inner-products, which are generally tied to the embedding dimension. Given the importance of embedding dimension for retrieval capacity, understanding how dense retrieval performance changes as embedding dimension is scaled is fundamental to building next-generation retrieval models that balance effectiveness and efficiency. In this work, we conduct a comprehensive analysis of the relationship between embedding dimension and retrieval performance. Our experiments include a range of model sizes from two model families to construct a detailed picture of embedding scaling behavior. We find that the scaling behavior fits a power law, allowing us to derive scaling laws for performance given only the embedding dimension, as well as a joint law accounting for embedding dimension and model size. Our analysis shows that for evaluation tasks aligned with the training task, performance continues to improve as embedding size increases, though with diminishing returns. For evaluation data that are less aligned with the training task, we find that performance is less predictable, with performance degrading with larger embedding dimensions for certain tasks. We hope our work provides additional insight into the limitations of embeddings and their behavior as well as offers a practical guide for selecting model and embedding dimension to achieve optimal performance with reduced storage and compute costs.

Modern information access ecosystems consist of mixtures of systems, such as retrieval systems and large language models, and increasingly rely on marketplaces to mediate access to models, tools, and data, making competition between systems inherent to deployment. In such settings, outcomes are shaped not only by benchmark quality but also by competitive pressure, including user switching, routing decisions, and operational constraints. Yet evaluation is still largely conducted on static benchmarks with accuracy-focused measures that assume systems operate in isolation. This mismatch makes it difficult to predict post-deployment success and obscures competitive effects such as early-adoption advantages and market dominance. We introduce Marketplace Evaluation, a simulation-based paradigm that evaluates information access systems as participants in a competitive marketplace. By simulating repeated interactions and evolving user and agent preferences, the framework enables longitudinal evaluation and marketplace-level metrics, such as retention and market share, that complement and can extend beyond traditional accuracy-based metrics. We formalize the framework and outline a research agenda, motivated by business and economics, around marketplace simulation, metrics, optimization, and adoption in evaluation campaigns like TREC.

Deep research agents have emerged as LLM-based systems designed to perform multi-step information seeking and reasoning over large, open-domain sources to answer complex questions by synthesizing information from multiple information sources. Given the complexity of the task and despite various recent efforts, evaluation of deep research agents remains fundamentally challenging. This paper identifies a list of requirements and optional properties for evaluating deep research agents. We observe that existing benchmarks do not satisfy all identified requirements. Inspired by prior research on TREC Total Recall Tracks, we introduce the task of Total Recall Question Answering and develop a framework for deep research agents evaluation that satisfies the identified criteria. Our framework constructs single-answer, total recall queries with precise evaluation and relevance judgments derived from a structured knowledge base paired with a text corpus, enabling large-scale data construction. Using this framework, we build TRQA, a deep research benchmark constructed from Wikidata-Wikipedia as a real-world source and a synthetically generated e-commerce knowledge base and corpus to mitigate the effects of data contamination. We benchmark the collection with representative retriever and deep research models and establish baseline retrieval and end-to-end results for future comparative evaluation.

Personalization is crucial for enhancing both the effectiveness and user satisfaction of language technologies, particularly in information-seeking tasks like question answering. Current approaches for personalizing large language models (LLMs) often rely on retrieval-augmented generation (RAG), followed by reinforcement learning with scalar reward signals to teach models how to use retrieved personal context. We believe that these scalar rewards sometimes provide weak, non-instructive feedback, limiting learning efficiency and personalization quality. We introduce VAC, a novel framework for personalized response generation that replaces scalar rewards with natural language feedback (NLF) that are generated conditioned on the user profiles and the question narratives. NLF serves as a rich and actionable supervision signal, allowing the policy model to iteratively refine its outputs and internalize effective personalization strategies. Training alternates between optimizing the feedback model and fine-tuning the policy model on the improved responses, resulting in a policy model that no longer requires feedback at inference. Evaluation on the LaMP-QA benchmark that consists of three diverse domains demonstrates consistent and significant improvements over the state-of-the-art results. Human evaluations further confirm the superior quality of the generated responses. These results demonstrate that NLF provides more effective signals for optimizing personalized question answering.

Retrieval-augmented reasoning (RAR) is a recent evolution of retrieval-augmented generation (RAG) that employs multiple reasoning steps for retrieval and generation. While effective for some complex queries, RAR remains vulnerable to errors and misleading outputs. Uncertainty quantification (UQ) offers methods to estimate the confidence of systems' outputs. These methods, however, often handle simple queries with no retrieval or single-step retrieval, without properly handling RAR setup. Accurate estimation of UQ for RAR requires accounting for all sources of uncertainty, including those arising from retrieval and generation. In this paper, we account for these sources and introduce Retrieval-Augmented Reasoning Consistency (R2C), a novel UQ method for RAR. The core idea of R2C is to perturb the multi-step reasoning process by applying various actions to reasoning steps. These perturbations alter the retriever's input, which shifts its output and consequently modifies the generator's input at the next step. Through this iterative feedback loop, the retriever and generator continuously reshape each other's inputs, enabling us to capture uncertainty arising from both components. Experiments on five popular RAR systems across diverse QA datasets show that R2C improves AUROC by over 5% on average compared to the state-of-the-art UQ baselines. Extrinsic evaluations using R2C as an external signal further confirm its effectiveness for two downstream tasks: in the Abstention task, it achieves ~5% gains in both F1Abstain and AccAbstain; in Model Selection, it improves exact match by ~7% over single models and ~3% over selection methods. Code is available on https://github.com/HeydarSoudani/R2C.

Personalized long-form question answering aims to tailor responses based on a user's historical data, which often spans diverse information domains. While prior work typically assumes user profiles are domain-aligned with the users' questions, real-world profiles are often diverse, containing information across multiple domains. In this work, we use the LaMP-QA benchmark, where user profiles span diverse Stack Exchange domains (Arts & Entertainment, Lifestyle & Personal Development, and Society & Culture), to systematically investigate how domain composition of the user profile affects personalization performance. We examine three profile composition settings: Cross-Domain, where profiles contain only out-of-domain information; Multi-Domain, where profiles span multiple domains including the question's domain; and In-Domain, where profiles match the question domain. Our analysis reveals that state-of-the-art personalization methods struggle in Cross-Domain settings, often failing to outperform non-personalized baselines. To address this challenge, we introduce Dual Retrieval Augmentation, a framework that jointly retrieves from both the user profile and a public corpus. We apply our framework to existing personalization approaches and demonstrate that it consistently outperforms state-of-the-art methods across all three profile settings, validating its robustness for personalized question answering.

Despite extensive research on a wide range of question answering (QA) systems, most existing work focuses on answer containment-i.e., assuming that answers can be directly extracted and/or generated from documents in the corpus. However, some questions require inference, i.e., deriving answers that are not explicitly stated but can be inferred from the available information. We introduce Inferential QA -- a new task that challenges models to infer answers from answer-supporting passages which provide only clues. To study this problem, we construct QUIT (QUestions requiring Inference from Texts) dataset, comprising 7,401 questions and 2.4M passages built from high-convergence human- and machine-authored hints, labeled across three relevance levels using LLM-based answerability and human verification. Through comprehensive evaluation of retrievers, rerankers, and LLM-based readers, we show that methods effective on traditional QA tasks struggle in inferential QA: retrievers underperform, rerankers offer limited gains, and fine-tuning provides inconsistent improvements. Even reasoning-oriented LLMs fail to outperform smaller general-purpose models. These findings reveal that current QA pipelines are not yet ready for inference-based reasoning. Inferential QA thus establishes a new class of QA tasks that move towards understanding and reasoning from indirect textual evidence.

Personalization is well studied in search and recommendation, but personalized question answering remains underexplored due to challenges in inferring preferences from long, noisy, implicit contexts and generating responses that are both accurate and aligned with user expectations. To address this, we propose Pathways of Thoughts (PoT), an inference-stage method that applies to any large language model (LLM) without task-specific fine-tuning. PoT models the thinking as an iterative decision process, where the model dynamically selects among cognitive operations such as reasoning, revision, personalization, and clarification. This enables exploration of multiple reasoning trajectories, producing diverse candidate responses that capture different perspectives. PoT then aggregates and reweights these candidates according to inferred user preferences, yielding a final personalized response that benefits from the complementary strengths of diverse reasoning paths. Experiments on the LaMP-QA benchmark show that PoT consistently outperforms competitive baselines, achieving up to a 10.8% relative improvement. Human evaluation further validates these improvements, with annotators preferring PoT in 66% of cases compared to the best-performing baseline and reporting ties in 15% of cases.

Large Language Models (LLMs) have transformed information access by enabling human-like text understanding and generation. This workshop explores the next step for conversational AI: building proactive information-seeking assistants that go beyond reactive question answering. We aim to investigate how LLMs can anticipate user needs, model complex context, support mixed-initiative interactions, integrate retrieval and external tools, personalize responses, adapt through feedback, and ensure fairness, transparency, and cognitive grounding. Bringing together experts from NLP, IR, HCI, and cognitive science, the workshop will serve as a timely forum for advancing intelligent, proactive dialogue systems. It will also foster interdisciplinary collaboration.

Retrieval-Enhanced Machine Learning (REML) refers to the use of information retrieval (IR) methods to support reasoning and inference in machine learning tasks. Although relatively recent, these approaches can substantially improve model performance. This includes improved generalization, knowledge grounding, scalability, freshness, attribution, interpretability, and on-device learning. To date, despite being influenced by work in the information retrieval community, REML research has predominantly been presented in natural language processing (NLP) conferences. Our tutorial addresses this disconnect by introducing core REML concepts and synthesizing the literature from various domains in machine learning (ML), including, but not limited to, NLP. What is unique to our approach is the use of consistent notations to provide researchers with a unified and expandable framework. The tutorial will be presented in lecture format based on an existing manuscript, with supporting materials and a comprehensive reading list available at a website. Building on the momentum of our successful workshop at SIGIR 2023 and our tutorial at SIGIR-AP 2024, this year's tutorial features updated content with an emphasis on retrieval technologies used across the broader ML community. We also highlight their role in emerging, future-facing applications such as language agents and evolving scenarios where the extensive body of knowledge from IR can provide critical insights and capabilities.

Existing information retrieval systems are largely constrained by their reliance on vector inner products to assess query-document relevance, which naturally limits the expressiveness of the relevance score they can produce.We propose a new paradigm; instead of representing a query as a vector, we use a small neural network that acts as a learned query-specific relevance function.This small neural network takes a document representation as input (in this work we use a single vector) and produces a scalar relevance score.To produce the small neural network we use a hypernetwork, a network that produces the weights of other networks, as our query encoder.We name this category of encoder models Hypencoders.Experiments on in-domain search tasks show that Hypencoders significantly outperform strong dense retrieval models and even surpass reranking models and retrieval models with an order of magnitude more parameters.To assess the extent of Hypencoders' capabilities, we evaluate on a set of hard retrieval tasks including tipof-the-tongue and instruction-following retrieval tasks.On harder tasks, we find that the performance gap widens substantially compared to standard retrieval tasks.Furthermore, to demonstrate the practicality of our method, we implement an approximate search algorithm and show that our model is able to retrieve from a corpus of 8.8M documents in under 60 milliseconds.

Personalized search is a problem where models benefit from learning user preferences from per-user historical interaction data. The inferred preferences enable personalized ranking models to improve the relevance of documents to users. However, personalization is also seen as opaque in its use of historical interactions and is not amenable to users' control. Further, personalization limits the diversity of information users are exposed to. While search results may be automatically diversified this does little to address the lack of control over personalization. In response, we introduce a model for personalized search that enables users to control personalized rankings proactively. Our model, CtrlCE, is a novel cross-encoder model augmented with an editable memory built from users' historical interactions. The editable memory allows cross-encoders to be personalized efficiently and enables users to control personalized ranking. Next, because all queries do not require personalization, we introduce a calibrated mixing model which determines when personalization is necessary. This enables users to control personalization via their editable memory only when necessary. To thoroughly evaluate CtrlCE, we demonstrate its empirical performance in four domains of science, its ability to selectively request user control in a calibration evaluation of the mixing model, and the control provided by its editable memory in a user study.

Scaling large language models (LLMs) has shown great potential for improving retrieval model performance; however, previous studies have mainly focused on dense retrieval trained with contrastive loss (CL), neglecting the scaling behavior of other retrieval paradigms and optimization techniques, such as sparse retrieval and knowledge distillation (KD). In this work, we conduct a systematic comparative study on how different retrieval paradigms (sparse vs. dense) and fine-tuning objectives (CL vs. KD vs. their combination) affect retrieval performance across different model scales. Using MSMARCO passages as the training dataset, decoder-only LLMs (Llama-3 series: 1B, 3B, 8B), and a fixed compute budget, we evaluate various training configurations on both in-domain (MSMARCO, TREC DL) and out-of-domain (BEIR) benchmarks. Our key findings reveal that: (1) Scaling behaviors emerge clearly only with CL, where larger models achieve significant performance gains, whereas KD-trained models show minimal improvement, performing similarly across the 1B, 3B, and 8B scales. (2) Sparse retrieval models consistently outperform dense retrieval across both in-domain (MSMARCO, TREC DL) and out-of-domain (BEIR) benchmarks, and they demonstrate greater robustness to imperfect supervised signals. (3) We successfully scale sparse retrieval models with the combination of CL and KD losses at 8B scale, achieving state-of-the-art (SOTA) results in all evaluation sets.

The effectiveness of clarification question models in engaging users within search systems is currently constrained, casting doubt on their overall usefulness. To improve the performance of these models, it is crucial to employ assessment approaches that encompass both real-time feedback from users (online evaluation) and the characteristics of clarification questions evaluated through human assessment (offline evaluation). However, the relationship between online and offline evaluations has been debated in information retrieval. This study aims to investigate how this discordance holds in search clarification. We use user engagement as ground truth and employ several offline labels to investigate to what extent the offline ranked lists of clarification resemble the ideal ranked lists based on online user engagement. Contrary to the current understanding that offline evaluations fall short of supporting online evaluations, we indicate that when identifying the most engaging clarification questions from the user’s perspective, online and offline evaluations correspond with each other. We show that the query length does not influence the relationship between online and offline evaluations, and reducing uncertainty in online evaluation strengthens this relationship. We illustrate that an engaging clarification needs to excel from multiple perspectives, and SERP quality and characteristics of the clarification are equally important. We also investigate if human labels can enhance the performance of Large Language Models (LLMs) and Learning-to-Rank (LTR) models in identifying the most engaging clarification questions from the user’s perspective by incorporating offline evaluations as input features. Our results indicate that LTR models do not perform better than individual offline labels. However, GPT, an LLM, emerges as the standout performer, surpassing all LTR models and offline labels.

Personalization is key in understanding user behavior and has been a main focus in the fields of knowledge discovery and information retrieval. Building personalized recommender systems is especially important now due to the vast amount of user-generated textual content, which offers deep insights into user preferences. The recent advancements in Large Language Models (LLMs) have significantly impacted research areas, mainly in Natural Language Processing and Knowledge Discovery, giving these models the ability to handle complex tasks and learn context. However, the use of generative models and user-generated text for personalized systems and recommendation is relatively new and has shown some promising results. This workshop is designed to bridge the research gap in these fields and explore personalized applications and recommender systems. We aim to fully leverage generative models to develop AI systems that are not only accurate but also focused on meeting individual user needs. Building upon the momentum of previous successful forums, this workshop seeks to engage a diverse audience from academia and industry, fostering a dialogue that incorporates fresh insights and anticipates over 50 attendees, including key stakeholders in the field.

This is the second workshop in the series which discusses the cutting-edge developments in research and applications of personalizing large language models (LLMs) and adapting them to the demands of diverse user populations and societal needs. The full-day workshop plan includes several keynotes and invited talks, a poster session and a panel discussion.

This paper studies retrieval-augmented approaches for personalizing large language models (LLMs), which potentially have a substantial impact on various applications and domains. We propose the first attempt to optimize the retrieval models that deliver a limited number of personal documents to large language models for the purpose of personalized generation. We develop two optimization algorithms that solicit feedback from the downstream personalized generation tasks for retrieval optimization--one based on reinforcement learning whose reward function is defined using any arbitrary metric for personalized generation and another based on knowledge distillation from the downstream LLM to the retrieval model. This paper also introduces a pre- and post-generation retriever selection model that decides what retriever to choose for each LLM input. Extensive experiments on diverse tasks from the language model personalization (LaMP) benchmark reveal statistically significant improvements in six out of seven datasets.

This paper introduces uRAG-a framework with a unified retrieval engine that serves multiple downstream retrieval-augmented generation (RAG) systems. Each RAG system consumes the retrieval results for a unique purpose, such as open-domain question answering, fact verification, entity linking, and relation extraction. We introduce a generic training guideline that standardizes the communication between the search engine and the downstream RAG systems that engage in optimizing the retrieval model. This lays the groundwork for us to build a large-scale experimentation ecosystem consisting of 18 RAG systems that engage in training and 18 unknown RAG systems that use the uRAG as the new users of the search engine. Using this experimentation ecosystem, we answer a number of fundamental research questions that improve our understanding of promises and challenges in developing search engines for machines.

Evaluating retrieval-augmented generation (RAG) presents challenges, particularly for retrieval models within these systems. Traditional end-to-end evaluation methods are computationally expensive. Furthermore, evaluation of the retrieval model's performance based on query-document relevance labels shows a small correlation with the RAG system's downstream performance. We propose a novel evaluation approach, eRAG, where each document in the retrieval list is individually utilized by the large language model within the RAG system. The output generated for each document is then evaluated based on the downstream task ground truth labels. In this manner, the downstream performance for each document serves as its relevance label. We employ various downstream task metrics to obtain document-level annotations and aggregate them using set-based or ranking metrics. Extensive experiments on a wide range of datasets demonstrate that eRAG achieves a higher correlation with downstream RAG performance compared to baseline methods, with improvements in Kendall's tau correlation ranging from 0.168 to 0.494. Additionally, eRAG offers significant computational advantages, improving runtime and consuming up to 50 times less GPU memory than end-to-end evaluation.

The field of conversational information seeking, which is rapidly gaining interest in both academia and industry, is changing how we interact with search engines through natural language interactions. Existing datasets and methods are mostly evaluating reactive conversational information seeking systems that solely provide response to every query from the user. We identify a gap in building and evaluating proactive conversational information seeking systems that can monitor a multi-party human conversation and proactively engage in the conversation at an opportune moment by retrieving useful resources and suggestions. In this paper, we introduce a large-scale dataset for proactive document retrieval that consists of over 2.8 million conversations. We conduct crowdsourcing experiments to obtain high-quality and relatively complete relevance judgments through depth-k pooling. We also collect annotations related to the parts of the conversation that are related to each document, enabling us to evaluate proactive retrieval systems. We introduce normalized proactive discounted cumulative gain (npDCG) for evaluating these systems, and further provide benchmark results for a wide range of models, including a novel model we developed for this task. We believe that the developed dataset, called ProCIS, paves the path towards developing proactive conversational information seeking systems.

This paper introduces Stochastic RAG--a novel approach for end-to-end optimization of retrieval-augmented generation (RAG) models that relaxes the simplifying assumptions of marginalization and document independence, made in most prior work. Stochastic RAG casts the retrieval process in RAG as a stochastic sampling without replacement process. Through this formulation, we employ straight-through Gumbel-top-k that provides a differentiable approximation for sampling without replacement and enables effective end-to-end optimization for RAG. We conduct extensive experiments on seven diverse datasets on a wide range of tasks, from open-domain question answering to fact verification to slot-filling for relation extraction and to dialogue systems. By applying this optimization method to a recent and effective RAG model, we advance state-of-the-art results on six out of seven datasets.

This paper introduces PAG-a novel optimization and decoding approach that guides autoregressive generation of document identifiers in generative retrieval models through simultaneous decoding. To this aim, PAG constructs a set-based and sequential identifier for each document. Motivated by the bag-of-words assumption in information retrieval, the set-based identifier is built on lexical tokens. The sequential identifier, on the other hand, is obtained via quantizing relevance-based representations of documents. Extensive experiments on MSMARCO and TREC Deep Learning Track data reveal that PAG outperforms the state-of-the-art generative retrieval model by a large margin (e.g., 15.6% MRR improvements on MS MARCO), while achieving 22× speed up in terms of query latency.

This workshop discusses the cutting-edge developments in research and applications of personalizing large language models (LLMs) and adapting them to the demands of diverse user populations and societal needs. The full-day workshop includes several keynotes and invited talks, a poster session and a panel discussion.

Over the past few years, consumer behavior has shifted from traditional in-store shopping to online shopping. For example, eCommerce sales have grown from around 5% of total US sales in 2012 to around 15.4% in year 2023. This rapid growth of eCommerce has created new challenges and vital new requirements for intelligent information retrieval systems. Which lead to the primary motivations of this workshop:

Recent research has shown that transformer networks can be used as differentiable search indexes by representing each document as a sequence of document ID tokens. These generative retrieval models cast the retrieval problem to a document ID generation problem for each query. Despite their elegant design, existing generative retrieval models only perform well on artificially-constructed and small-scale collections. This paper represents an important milestone in generative retrieval research by showing that generative retrieval models can be trained to perform effectively on large-scale standard retrieval benchmarks. In more detail, we propose RIPOR- an optimization framework for generative retrieval that is designed based on two often-overlooked fundamental design considerations. First, RIPOR introduces a novel prefix-oriented ranking optimization algorithm for accurate estimation of relevance score during sequential document ID generation. Second, RIPOR constructs document IDs based on the relevance associations between queries and documents. Evaluation on MSMARCO and TREC Deep Learning Track reveals that RIPOR surpasses state-of-the-art generative retrieval models by a large margin (e.g., 30.5% MRR improvements on MS MARCO Dev Set).

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.

In this article, we provide an overview of ACM TWEB’s Special Issue on Conversational Information Seeking. It highlights both research and practical applications in this field. The article also discusses the future potential of conversational information seeking technology.

Narrative-driven recommendation (NDR) presents an information access problem where users solicit recommendations with verbose descriptions of their preferences and context, for example, travelers soliciting recommendations for points of interest while describing their likes/dislikes and travel circumstances. These requests are increasingly important with the rise of natural language-based conversational interfaces for search and recommendation systems. However, NDR lacks abundant training data for models, and current platforms commonly do not support these requests. Fortunately, classical user-item interaction datasets contain rich textual data, e.g., reviews, which often describe user preferences and context – this may be used to bootstrap training for NDR models. In this work, we explore using large language models (LLMs) for data augmentation to train NDR models. We use LLMs for authoring synthetic narrative queries from user-item interactions with few-shot prompting and train retrieval models for NDR on synthetic queries and user-item interaction data. Our experiments demonstrate that this is an effective strategy for training small-parameter retrieval models that outperform other retrieval and LLM baselines for narrative-driven recommendation.

Most machine learning models are designed to be self-contained and encode both "knowledge" and "reasoning" in their parameters. However, such models cannot perform effectively for tasks that require knowledge grounding and tasks that deal with non-stationary data, such as news and social media. Besides, these models often require huge number of parameters to encode all the required knowledge. These issues can be addressed via augmentation with a retrieval model. This category of machine learning models, which is called Retrieval-enhanced machine learning (REML), has recently attracted considerable attention in multiple research communities. For instance, REML models have been studied in the context of open-domain question answering, fact verification, and dialogue systems and also in the context of generalization through memorization in language models and memory networks. We believe that the information retrieval community can significantly contribute to this growing research area by designing, implementing, analyzing, and evaluating various aspects of retrieval models with applications to REML tasks. The goal of this full-day hybrid workshop is to bring together researchers from industry and academia to discuss various aspects of retrieval-enhanced machine learning, including effectiveness, efficiency, and robustness of these models in addition to their impact on real-world applications.

Retrieval-augmented generation models offer many benefits over standalone language models: besides a textual answer to a given query they provide provenance items retrieved from an updateable knowledge base. However, they are also more complex systems and need to handle long inputs. In this work, we introduce FiD-Light to strongly increase the efficiency of the state-of-the-art retrieval-augmented FiD model, while maintaining the same level of effectiveness. Our FiD-Light model constrains the information flow from the encoder (which encodes passages separately) to the decoder (using concatenated encoded representations). Furthermore, we adapt FiD-Light with re-ranking capabilities through textual source pointers, to improve the top-ranked provenance precision. Our experiments on a diverse set of seven knowledge intensive tasks (KILT) show FiD-Light consistently improves the Pareto frontier between query latency and effectiveness. FiD-Light with source pointing sets substantial new state-of-the-art results on six KILT tasks for combined text generation and provenance retrieval evaluation, while maintaining high efficiency.

In this work, we explore a Multilingual Information Retrieval (MLIR) task, where the collection includes documents in multiple languages. We demonstrate that applying state-of-the-art approaches developed for cross-lingual information retrieval to MLIR tasks leads to sub-optimal performance. This is due to the heterogeneous and imbalanced nature of multilingual collections -- some languages are better represented in the collection and some benefit from large-scale training data. To address this issue, we present KD-SPD, a novel soft prompt decoding approach for MLIR that implicitly "translates'' the representation of documents in different languages into the same embedding space. To address the challenges of data scarcity and imbalance, we introduce a knowledge distillation strategy. The teacher model is trained on rich English retrieval data, and by leveraging bi-text data, our distillation framework transfers its retrieval knowledge to the multilingual document encoder. Therefore, our approach does not require any multilingual retrieval training data. Extensive experiments on three MLIR datasets with a total of 15 languages demonstrate that KD-SPD significantly outperforms competitive baselines in all cases. We conduct extensive analyses to show that our method has less language bias and better zero-shot transfer ability towards new languages.

Methods for making high-quality recommendations often rely on learning latent representations from interaction data. These methods, while performant, do not provide ready mechanisms for users to control the recommendation they receive. Our work tackles this problem by proposing LACE, a novel concept value bottleneck model for controllable text recommendations. LACE represents each user with a succinct set of human-readable concepts through retrieval given user-interacted documents and learns personalized representations of the concepts based on user documents. This concept based user profile is then leveraged to make recommendations. The design of our model affords control over the recommendations through a number of intuitive interactions with a transparent user profile. We first establish the quality of recommendations obtained from LACE in an offline evaluation on three recommendation tasks spanning six datasets in warm-start, cold-start, and zero-shot setups. Next, we validate the controllability of LACE under simulated user interactions. Finally, we implement LACE in an interactive controllable recommender system and conduct a user study to demonstrate that users are able to improve the quality of recommendations they receive through interactions with an editable user profile.

Knowledge-Intensive Visual Question Answering (KI-VQA) refers to answering a question about an image whose answer does not lie in the image. This paper presents a new pipeline for KI-VQA tasks, consisting of a retriever and a reader. First, we introduce DEDR, a symmetric dual encoding dense retrieval framework in which documents and queries are encoded into a shared embedding space using uni-modal (textual) and multi-modal encoders. We introduce an iterative knowledge distillation approach that bridges the gap between the representation spaces in these two encoders. Extensive evaluation on two well-established KI-VQA datasets, i.e., OK-VQA and FVQA, suggests that DEDR outperforms state-of-the-art baselines by 11.6% and 30.9% on OK-VQA and FVQA, respectively.

Dense retrieval models use bi-encoder network architectures for learning query and document representations. These representations are often in the form of a vector representation and their similarities are often computed using the dot product function. In this paper, we propose a new representation learning framework for dense retrieval. Instead of learning a vector for each query and document, our framework learns a multivariate distribution and uses negative multivariate KL divergence to compute the similarity between distributions. For simplicity and efficiency reasons, we assume that the distributions are multivariate normals and then train large language models to produce mean and variance vectors for these distributions. We provide a theoretical foundation for the proposed framework and show that it can be seamlessly integrated into the existing approximate nearest neighbor algorithms to perform retrieval efficiently. We conduct an extensive suite of experiments on a wide range of datasets, and demonstrate significant improvements compared to competitive dense retrieval models.

Developing a universal model that can efficiently and effectively respond to a wide range of information access requests-from retrieval to recommendation to question answering---has been a long-lasting goal in the information retrieval community. This paper argues that the flexibility, efficiency, and effectiveness brought by the recent development in dense retrieval and approximate nearest neighbor search have smoothed the path towards achieving this goal. We develop a generic and extensible dense retrieval framework, called framework, that can handle a wide range of (personalized) information access requests, such as keyword search, query by example, and complementary item recommendation. Our proposed approach extends the capabilities of dense retrieval models for ad-hoc retrieval tasks by incorporating user-specific preferences through the development of a personalized attentive network. This allows for a more tailored and accurate personalized information access experience. Our experiments on real-world e-commerce data suggest the feasibility of developing universal information access models by demonstrating significant improvements even compared to competitive baselines specifically developed for each of these individual information access tasks. This work opens up a number of fundamental research directions for future exploration.

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.

Conversational information seeking (CIS) involves interaction sequences between one or more users and an information system. Interactions in CIS are primarily based on natural language dialogue, while they may include other types of interactions, such as click, touch, and body gestures. CIS recently attracted significant attention and advancements continue to be made. This tutorial follows the content of the recent Conversational Information Seeking book authored by several of the tutorial presenters. The tutorial aims to be an introduction to CIS for newcomers to CIS in addition to the recent advanced topics and state-of-the-art approaches for students and researchers with moderate knowledge of the topic. A significant part of the tutorial is dedicated to hands-on experiences based on toolkits developed by the presenters for conversational passage retrieval and multi-modal task-oriented dialogues. The outcomes of this tutorial include theoretical and practical knowledge, including a forum to meet researchers interested in CIS.

Asking clarification questions is an active area of research; however, resources for training and evaluating search clarification methods are not sufficient. To address this issue, we describe MIMICS-Duo, a new freely available dataset of 306 search queries with multiple clarifications (a total of 1,034 query-clarification pairs). MIMICS-Duo contains fine-grained annotations on clarification questions and their candidate answers and enhances the existing MIMICS datasets by enabling multi-dimensional evaluation of search clarification methods, including online and offline evaluation. We conduct extensive analysis to demonstrate the relationship between offline and online search clarification datasets and outline several research directions enabled by MIMICS-Duo. We believe that this resource will help researchers better understand clarification in search.

Although information access systems have long supportedpeople in accomplishing a wide range of tasks, we propose broadening the scope of users of information access systems to include task-driven machines, such as machine learning models. In this way, the core principles of indexing, representation, retrieval, and ranking can be applied and extended to substantially improve model generalization, scalability, robustness, and interpretability. We describe a generic retrieval-enhanced machine learning (REML) framework, which includes a number of existing models as special cases. REML challenges information retrieval conventions, presenting opportunities for novel advances in core areas, including optimization. The REML research agenda lays a foundation for a new style of information access research and paves a path towards advancing machine learning and artificial intelligence.

Recent work has shown that more effective dense retrieval models can be obtained by distilling ranking knowledge from an existing base re-ranking model. In this paper, we propose a generic curriculum learning based optimization framework called CL-DRD that controls the difficulty level of training data produced by the re-ranking (teacher) model. CL-DRD iteratively optimizes the dense retrieval (student) model by increasing the difficulty of the knowledge distillation data made available to it. In more detail, we initially provide the student model coarse-grained preference pairs between documents in the teacher's ranking, and progressively move towards finer-grained pairwise document ordering requirements. In our experiments, we apply a simple implementation of the CL-DRD framework to enhance two state-of-the-art dense retrieval models. Experiments on three public passage retrieval datasets demonstrate the effectiveness of our proposed framework.

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.

While current information retrieval systems are effective for known-item retrieval where the searcher provides a precise name or identifier for the item being sought, systems tend to be much less effective for cases where the searcher is unable to express a precise name or identifier. We refer to this as tip of the tongue (TOT) known-item retrieval, named after the cognitive state of not being able to retrieve an item from memory. Using movie search as a case study, we explore the characteristics of questions posed by searchers in TOT states in a community question answering website. We analyze how searchers express their information needs during TOT states in the movie domain. Specifically, what information do searchers remember about the item being sought and how do they convey this information? Our results suggest that searchers use a combination of information about: (1) the content of the item sought, (2) the context in which they previously engaged with the item, and (3) previous attempts to find the item using other resources (e.g., search engines). Additionally, searchers convey information by sometimes expressing uncertainty (i.e., hedging), opinions, emotions, and by performing relative (vs. absolute) comparisons with attributes of the item. As a result of our analysis, we believe that searchers in TOT states may require specialized query understanding methods or document representations. Finally, our preliminary retrieval experiments show the impact of each information type presented in information requests on retrieval performance.

Information seeking conversations between users and Conversational Search Agents (CSAs) consist of multiple turns of interaction. While users initiate a search session, ideally a CSA should sometimes take the lead in the conversation by obtaining feedback from the user by offering query suggestions or asking for query clarifications i.e. mixed initiative. This creates the potential for more engaging conversational searches, but substantially increases the complexity of modelling and evaluating such scenarios due to the large interaction space coupled with the trade-offs between the costs and benefits of the different interactions. In this paper, we present a model for conversational search -- from which we instantiate different observed conversational search strategies, where the agent elicits: (i) Feedback-First, or (ii) Feedback-After. Using 49 TREC WebTrack Topics, we performed an analysis comparing how well these different strategies combine with different mixed initiative approaches: (i) Query Suggestions vs. (ii) Query Clarifications. Our analysis reveals that there is no superior or dominant combination, instead it shows that query clarifications are better when asked first, while query suggestions are better when asked after presenting results. We also show that the best strategy and approach depends on the trade-offs between the relative costs between querying and giving feedback, the performance of the initial query, the number of assessments per query, and the total amount of gain required. While this work highlights the complexities and challenges involved in analyzing CSAs, it provides the foundations for evaluating conversational strategies and conversational search agents in batch/offline settings.

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.

Recent research on conversational information seeking (CIS) mostly focuses on uni-modal interactions and information items. This per- spective paper highlights the importance of moving towards de- veloping and evaluating multi-modal conversational information seeking (MMCIS) systems as they enable us to leverage richer context, overcome errors, and increase accessibility. We bridge the gap between the multi-modal and CIS research and provide a formal definition for MMCIS. We discuss potential opportunities and research challenges in designing, implementing, and evaluating MMCIS systems. Based on this research, we propose and implement a practical open-source framework for facilitating MMCIS research.

An emerging recipe for achieving state-of-the-art effectiveness in neural document re-ranking involves utilizing large pre-trained language models - e.g., BERT - to evaluate all individual passages in the document and then aggregating the outputs by pooling or additional Transformer layers. A major drawback of this approach is high query latency due to the cost of evaluating every passage in the document with BERT. To make matters worse, this high inference cost and latency varies based on the length of the document, with longer documents requiring more time and computation. To address this challenge, we adopt an intra-document cascading strategy, which prunes passages of a candidate document using a less expensive model, called ESM, before running a scoring model that is more expensive and effective, called ETM. We found it best to train ESM (short for Efficient Student Model) via knowledge distillation from the ETM (short for Effective Teacher Model) e.g., BERT. This pruning allows us to only run the ETM model on a smaller set of passages whose size does not vary by document length. Our experiments on the MS MARCO and TREC Deep Learning Track benchmarks suggest that the proposed Intra-Document Cascaded Ranking Model (IDCM) leads to over 400% lower query latency by providing essentially the same effectiveness as the state-of-the-art BERT-based document ranking models.

Podcasts are spoken documents across a wide-range of genres and styles, with growing listenership across the world, and a rapidly lowering barrier to entry for both listeners and creators. The great strides in search and recommendation in research and industry have yet to see impact in the podcast space, where recommendations are still largely driven by word of mouth. In this perspective paper, we highlight the many differences between podcasts and other media, and discuss our perspective on challenges and future research directions in the domain of podcast information access.

The Transformer-Kernel (TK) model has demonstrated strong reranking performance on the TREC Deep Learning benchmark---and can be considered to be an efficient (but slightly less effective) alternative to other Transformer-based architectures that employ (i) large-scale pretraining (high training cost), (ii) joint encoding of query and document (high inference cost), and (iii) larger number of Transformer layers (both high training and high inference costs). Since, a variant of the TK model---called TKL---has been developed that incorporates local self-attention to efficiently process longer input sequences in the context of document ranking. In this work, we propose a novel Conformer layer as an alternative approach to scale TK to longer input sequences. Furthermore, we incorporate query term independence and explicit term matching to extend the model to the full retrieval setting. We benchmark our models under the strictly blind evaluation setting of the TREC 2020 Deep Learning track and find that our proposed architecture changes lead to improved retrieval quality over TKL. Our best model also outperforms all non-neural runs ("trad") and two-thirds of the pretrained Transformer-based runs ("nnlm") on [email protected]

Recent deployment of efficient billion-scale approximate nearest neighbor (ANN) search algorithms on GPUs has motivated information retrieval researchers to develop neural ranking models that learn low-dimensional dense representations for queries and documents and use ANN search for retrieval. However, optimizing these dense retrieval models poses several challenges including negative sampling for (pair-wise) training. A recent model, called ANCE, successfully uses dynamic negative sampling using ANN search. This paper improves upon ANCE by proposing a robust negative sampling strategy for scenarios where the training data lacks complete relevance annotations. This is of particular importance as obtaining large-scale training data with complete relevance judgment is extremely expensive. Our model uses a small validation set with complete relevance judgments to accurately estimate a negative sampling distribution for dense retrieval models. We also explore leveraging a lexical matching signal during training and pseudo-relevance feedback during evaluation for improved performance. Our experiments on the TREC Deep Learning Track benchmarks demonstrate the effectiveness of our solutions.

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.

Information Retrieval has a long history of applying either discriminative or generative modeling to retrieval and ranking tasks. Recent developments in transformer architectures and multi-task learning techniques have dramatically improved our ability to train effective neural models capable of resolving a wide variety of tasks using either of these paradigms. In this paper, we propose a novel multi-task learning approach which can be used to produce more effective neural ranking models. The key idea is to improve the quality of the underlying transformer model by cross-training a retrieval task and one or more complementary language generation tasks. By targeting the training on the encoding layer in the transformer architecture, our experimental results show that the proposed multi-task learning approach consistently improves retrieval effectiveness on the targeted collection and can easily be re-targeted to new ranking tasks. We provide an in-depth analysis showing how multi-task learning modifies model behaviors, resulting in more general models.

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.

An introduction to the special issue on conversational search and recommendation is presented in this article. While conversational search and recommendation has roots in early Information Retrieval (IR) research, the recent advances in automatic voice recognition and conversational agents have created increasing interest in this area. In recent years, the IR and related communities have witnessed a number of major contributions to the field of conversational search and recommendation. They include but are not limited to conversational search conceptualization. The growing body of work in this area has been supplemented by an increasing number of recent seminars.

Search clarification has recently attracted much attention due to its applications in search engines. It has also been recognized as a major component in conversational information seeking systems. Despite its importance, the research community still feels the lack of a large-scale dataset for studying different aspects of search clarification. In this paper, we introduce MIMICS, a collection of search clarification datasets for real web search queries sampled from the Bing query logs. Each clarification in MIMICS is generated by a Bing production algorithm and consists of a clarifying question and up to five candidate answers. MIMICS contains three datasets: (1) MIMICS-Click includes over 400k unique queries, their associated clarification panes, and the corresponding aggregated user interaction signals (i.e., clicks). (2) MIMICS-ClickExplore is an exploration data that includes aggregated user interaction signals for over 60k unique queries, each with multiple clarification panes. (3) MIMICS-Manual includes over 2k unique real search queries. Each query-clarification pair in this dataset has been manually labeled by at least three trained annotators. It contains graded quality labels for the clarifying question, the candidate answer set, and the landing result page for each candidate answer.

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.

Neural networks, particularly Transformer-based architectures, have achieved significant performance improvements on several retrieval benchmarks. When the items being retrieved are documents, the time and memory cost of employing Transformers over a full sequence of document terms can be prohibitive. A popular strategy involves considering only the first n terms of the document. This can, however, result in a biased system that under retrieves longer documents. In this work, we propose a local self-attention which considers a moving window over the document terms and for each term attends only to other terms in the same window. This local attention incurs a fraction of the compute and memory cost of attention over the whole document. The windowed approach also leads to more compact packing of padded documents in minibatches resulting in additional savings. We also employ a learned saturation function and a two-staged pooling strategy to identify relevant regions of the document. The Transformer-Kernel pooling model with these changes can efficiently elicit relevance information from documents with thousands of tokens. We benchmark our proposed modifications on the document ranking task from the TREC 2019 Deep Learning track and observe significant improvements in retrieval quality as well as increased retrieval of longer documents at moderate increase in compute and memory costs.

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.

We present RML, the first known general reinforcement learning framework for relevance feedback that directly optimizes any desired retrieval metric, including precision-oriented, recall-oriented, and even diversity metrics: RML can be easily extended to directly optimize any arbitrary user satisfaction signal. Using the RML framework, we can select effective feedback terms and weight them appropriately, improving on past methods that fit parameters to feedback algorithms using heuristic approaches or methods that do not directly optimize for retrieval performance. Learning an effective relevance feedback model is not trivial since the true feedback distribution is unknown. Experiments on standard TREC collections compare RML to existing feedback algorithms, demonstrate the effectiveness of RML at optimizing for MAP and α-n DCG, and show the impact on related measures.

Conversational information seeking (CIS) has been recognized as a major emerging research area in information retrieval. Such research will require data and tools, to allow the implementation and study of conversational systems. This paper introduces Macaw, an open-source framework with a modular architecture for CIS research. Macaw supports multi-turn, multi-modal, and mixed-initiative interactions, and enables research for tasks such as document retrieval, question answering, recommendation, and structured data exploration. It has a modular design to encourage the study of new CIS algorithms, which can be evaluated in batch mode. It can also integrate with a user interface, which allows user studies and data collection in an interactive mode, where the back end can be fully algorithmic or a wizard of oz setup. Macaw is distributed under the MIT License.

Asking clarifying questions in response to search queries has been recognized as a useful technique for revealing the underlying intent of the query. Clarification has applications in retrieval systems with different interfaces, from the traditional web search interfaces to the limited bandwidth interfaces as in speech-only and small screen devices. Generation and evaluation of clarifying questions have been recently studied in the literature. However, user interaction with clarifying questions is relatively unexplored. In this paper, we conduct a comprehensive study by analyzing large-scale user interactions with clarifying questions in a major web search engine. In more detail, we analyze the user engagements received by clarifying questions based on different properties of search queries, clarifying questions, and their candidate answers. We further study click bias in the data, and show that even though reading clarifying questions and candidate answers does not take significant efforts, there still exist some position and presentation biases in the data. We also propose a model for learning representation for clarifying questions based on the user interaction data as implicit feedback. The model is used for re-ranking a number of automatically generated clarifying questions for a given query. Evaluation on both click data and human labeled data demonstrates the high quality of the proposed method.

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.

Search queries are often short, and the underlying user intent may be ambiguous. This makes it challenging for search engines to predict possible intents, only one of which may pertain to the current user. To address this issue, search engines often diversify the result list and present documents relevant to multiple intents of the query. An alternative approach is to ask the user a question to clarify her information need. Asking clarifying questions is particularly important for scenarios with “limited bandwidth” interfaces, such as speech-only and small-screen devices. In addition, our user studies and large-scale online experiments show that asking clarifying questions is also useful in web search. Although some recent studies have pointed out the importance of asking clarifying questions, generating them for open-domain search tasks remains unstudied and is the focus of this paper. Lack of training data even within major search engines for this task makes it challenging. To mitigate this issue, we first identify a taxonomy of clarification for open-domain search queries by analyzing large-scale query reformulation data sampled from Bing search logs. This taxonomy leads us to a set of question templates and a simple yet effective slot filling algorithm. We further use this model as a source of weak supervision to automatically generate clarifying questions for training. Furthermore, we propose supervised and reinforcement learning models for generating clarifying questions learned from weak supervision data. We also investigate methods for generating candidate answers for each clarifying question, so users can select from a set of pre-defined answers. Human evaluation of the clarifying questions and candidate answers for hundreds of search queries demonstrates the effectiveness of the proposed solutions.

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).

With widespread use of mobile devices, instant messaging (IM) services have recently attracted a great deal of attention by millions of users. This has motivated news agencies to share their contents via such platforms in addition to their websites and popular social media. As a result, thousands of users nowadays follow the news agencies through their verified channels in IM services. However, user interactions with such platforms is relatively unstudied. In this paper, we provide an initial study to analyze and predict news popularity in an instant messaging service. To this aim, we focus on Telegram, a popular IM service with 200 million monthly active users. We explore the differences between news popularity analysis in Telegram and typical social media, such as Twitter, and highlight its unique characteristics. We perform our analysis on the data we collected from four diverse news agencies. Following our analysis, we study the task of news popularity prediction in Telegram and show that the performance of the prediction models can be substantially improved by learning from the data of multiple news agencies using multi-task learning. To foster research in this area, we have made the collected data publicly available.

The ACM Recommender Systems Challenge 2018 focused on the task of automatic music playlist continuation, which is a form of the more general task of sequential recommendation. Given a playlist of arbitrary length with some additional meta-data, the task was to recommend up to 500 tracks that fit the target characteristics of the original playlist. For the RecSys Challenge, Spotify released a dataset of one million user-generated playlists. Participants could compete in two tracks, i.e., main and creative tracks. Participants in the main track were only allowed to use the provided training set, however, in the creative track, the use of external public sources was permitted. In total, 113 teams submitted 1,228 runs to the main track; 33 teams submitted 239 runs to the creative track. The highest performing team in the main track achieved an R-precision of 0.2241, an NDCG of 0.3946, and an average number of recommended songs clicks of 1.784. In the creative track, an R-precision of 0.2233, an NDCG of 0.3939, and a click rate of 1.785 was obtained by the best team. This article provides an overview of the challenge, including motivation, task definition, dataset description, and evaluation. We further report and analyze the results obtained by the top-performing teams in each track and explore the approaches taken by the winners. We finally summarize our key findings, discuss generalizability of approaches and results to domains other than music, and list the open avenues and possible future directions in the area of automatic playlist continuation.

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.

The ACM Recommender Systems Challenge 2018 focused on automatic music playlist continuation, which is a form of the more general task of sequential recommendation. Given a playlist of arbitrary length, the challenge was to recommend up to 500 tracks that fit the target characteristics of the original playlist. For the Challenge, Spotify released a dataset of one million user-created playlists, along with associated metadata. Participants could submit their approaches in two tracks, i.e., main and creative tracks, where the former allowed teams to use solely the provided dataset and the latter allowed them to exploit publicly available external data too. In total, 113 teams submitted 1,228 runs in the main track; 33 teams submitted 239 runs in the creative track. The highest performing team in the main track achieved an R-precision of 0.2241, an NDCG of 0.3946, and an average number of recommended songs clicks of 1.784. In the creative track, an R-precision of 0.2233, an NDCG of 0.3939, and a click rate of 1.785 was realized by the best team.

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.

Does this sentence need citation? In this paper, we introduce the task of citation worthiness for scientific texts at a sentence-level granularity. The task is to detect whether a sentence in a scientific article needs to be cited or not. It can be incorporated into citation recommendation systems to help automate the citation process by marking sentences where needed. It may also be useful for publishers to regularize the citation process. We construct a dataset using the ACL Anthology Reference Corpus; consisting of over 1.1M "not_cite" and 85K "cite" sentences. We study the performance of a set of state-of-the-art sentence classifiers for the citation worthiness task and show the practical challenges. We also explore section-wise difficulty of the task and analyze the performance of our best model on a published article.

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.

Axiomatic analysis is a well-defined theoretical framework for analytical evaluation of information retrieval models. The current studies in axiomatic analysis implicitly assume that the constraints (axioms) are independent. In this paper, we revisit this assumption and hypothesize that there might be interdependence relationships between the existing constraints. As a preliminary study, we focus on the pseudo-relevance feedback (PRF) models that have been theoretically studied using the axiomatic analysis approach. In this paper, we introduce two novel interdependent PRF constraints which emphasize on the effect of existing constraints on each other. We further modify two state-of-the-art PRF models, log-logistic and relevance models, in order to satisfy the proposed constraints. Experiments on three TREC newswire and web collections demonstrate that the proposed modifications significantly outperform the baselines, in all cases.

In recent years, machine learning approaches, and in particular deep neural networks, have yielded significant improvements on several natural language processing and computer vision tasks; however, such breakthroughs have not yet been observed in the area of information retrieval. Besides the complexity of IR tasks, such as understanding the user's information needs, a main reason is the lack of high-quality and/or large-scale training data for many IR tasks. This necessitates studying how to design and train machine learning algorithms where there is no large-scale or high-quality data in hand. Therefore, considering the quick progress in development of machine learning models, this is an ideal time for a workshop that especially focuses on learning in such an important and challenging setting for IR tasks. The goal of this workshop is to bring together researchers from industry---where data is plentiful but noisy---with researchers from academia---where data is sparse but clean to discuss solutions to these related problems.

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.

Deep neural networks have recently shown promise in the ad-hoc retrieval task. However, such models have often been based on one field of the document, for example considering document title only or document body only. Since in practice documents typically have multiple fields, and given that non-neural ranking models such as BM25F have been developed to take advantage of document structure, this paper investigates how neural models can deal with multiple document fields. We introduce a model that can consume short text fields such as document title and long text fields such as document body. It can also handle multi-instance fields with variable number of instances, for example where each document has zero or more instances of incoming anchor text. Since fields vary in coverage and quality, we introduce a masking method to handle missing field instances, as well as a field-level dropout method to avoid relying too much on any one field. As in the studies of non-neural field weighting, we find it is better for the ranker to score the whole document jointly, rather than generate a per-field score and aggregate. We find that different document fields may match different aspects of the query and therefore benefit from comparing with separate representations of the query text. The combination of techniques introduced here leads to a neural ranker that can take advantage of full document structure, including multiple instance and missing instance data, of variable length. The techniques significantly enhance the performance of the ranker, and outperform a learning to rank baseline with hand-crafted features.

Content-based recommender systems (CBRSs) rely on user-item similarities that are calculated between user profiles and item representations. Appropriate representation of each user profile based on the user's past preferences can have a great impact on user's satisfaction in CBRSs. In this paper, we focus on text recommendation and propose a novel profile updating model based on previously recommended items as well as semantic similarity of terms calculated using distributed representation of words. We evaluate our model using two standard text recommendation datasets: TREC-9 Filtering Track and CLEF 2008-09 INFILE Track collections. Our experiments investigate the importance of both past recommended items and semantic similarities in recommendation performance. The proposed profile updating method significantly outperforms the baselines, which confirms the importance of incorporating semantic similarities in the profile updating task.

Number of terms in a query is a query-specific constant that is typically ignored in retrieval functions. However, previous studies have shown that the performance of retrieval models varies for different query lengths, and it usually degrades when query length increases. A possible reason for this issue can be the extraneous terms in longer queries that makes it a challenge for the retrieval models to distinguish between the key and complementary concepts of the query. As a signal to understand the importance of a term, inverse document frequency (IDF) can be used to discriminate query terms. In this paper, we propose a constraint to model the interaction between query length and IDF. Our theoretical analysis shows that current state-of-the-art retrieval models, such as BM25, do not satisfy the proposed constraint. We further analyze the BM25 model and suggest a modification to adapt BM25 so that it adheres to the new constraint. Our experiments on three TREC collections demonstrate that the proposed modification outperforms the baselines, especially for verbose queries.

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.

Pseudo-relevance feedback (PRF) refers to a query expansion strategy based on top-retrieved documents, which has been shown to be highly effective in many retrieval models. Previous work has introduced a set of constraints (axioms) that should be satisfied by any PRF model. In this paper, we propose three additional constraints based on the proximity of feedback terms to the query terms in the feedback documents. As a case study, we consider the log-logistic model, a state-of-the-art PRF model that has been proven to be a successful method in satisfying the existing PRF constraints, and show that it does not satisfy the proposed constraints. We further modify the log-logistic model based on the proposed proximity-based constraints. Experiments on four TREC collections demonstrate the effectiveness of the proposed constraints. Our modification the log-logistic model leads to significant and substantial (up to 15%) improvements. Furthermore, we show that the proposed proximity-based function outperforms the well-known Gaussian kernel which does not satisfy all the proposed constraints.

Exploitation of term relatedness provided by word embedding has gained considerable attention in recent IR literature. However, an emerging question is whether this sort of relatedness fits to the needs of IR with respect to retrieval effectiveness. While we observe a high potential of word embedding as a resource for related terms, the incidence of several cases of topic shifting deteriorates the final performance of the applied retrieval models. To address this issue, we revisit the use of global context (i.e. the term co-occurrence in documents) to measure the term relatedness. We hypothesize that in order to avoid topic shifting among the terms with high word embedding similarity, they should often share similar global contexts as well. We therefore study the effectiveness of post filtering of related terms by various global context relatedness measures. Experimental results show significant improvements in two out of three test collections, and support our initial hypothesis regarding the importance of considering global context in retrieval.

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.

Modern search engines leverage a variety of sources, beyond the conventional query-document content similarity, to improve their ranking performance. Among them, query context has attracted attention in prior work. Previously, query context was mainly modeled by user search history, either long-term or short-term, to help the ranking of future queries. In this paper, we focus on situational context, i.e., the contextual features of the current search request that are independent from both query content and user history. As an example, situational context can depend on search request time and location. We propose two context-aware ranking models based on neural networks. The first model learns a low-dimensional deep representation from the combination of contextual features. The second model extends the first one by leveraging binarized contextual features in addition to the high-level abstractions learned using a deep network.

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.

Pseudo-relevance feedback (PRF) has been proven to be an effective query expansion strategy to improve retrieval performance. Several PRF methods have so far been proposed for many retrieval models. Recent theoretical studies of PRF methods show that most of the PRF methods do not satisfy all necessary constraints. Among all, the log-logistic model has been shown to be an effective method that satisfies most of the PRF constraints. In this paper, we first introduce two new PRF constraints. We further analyze the log-logistic feedback model and show that it does not satisfy these two constraints as well as the previously proposed "relevance effect" constraint. We then modify the log-logistic formulation to satisfy all these constraints. Experiments on three TREC newswire and web collections demonstrate that the proposed modification significantly outperforms the original log-logistic model, in all collections.

User engagement evaluation task in social networks has recently attracted considerable attention due to its applications in recommender systems. In this task, the posts containing users' opinions about items, e.g., the tweets containing the users' ratings about movies in the IMDb website, are studied. In this paper, we try to make use of tweets from different web applications to improve the user engagement evaluation performance. To this aim, we propose an adaptive method based on multi-task learning. Since in this paper we study the problem of detecting tweets with positive engagement which is a highly imbalanced classification problem, we modify the loss function of multi-task learning algorithms to cope with the imbalanced data. Our evaluations over a dataset including the tweets of four diverse and popular data sources, i.e., IMDb, YouTube, Goodreads, and Pandora, demonstrate the effectiveness of the proposed method. Our findings suggest that transferring knowledge between data sources can improve the user engagement evaluation performance.