The exponential growth of scientific literature has created a pressing need for automated survey generation. Although recent LLM-based agents have shown promise in automating this task, current progress is hindered by the lack of a standardized, scalable evaluation protocol. Existing evaluation methods typically rely on either human evaluation or custom metrics designed to validate specific pipelines, which restricts scalability and hinders fair comparison. To address this, we introduce SurGE, a benchmark and evaluation framework tailored for scientific survey generation. SurGE provides a large-scale retrieval corpus of over one million papers and expert-validated ground-truth surveys. Furthermore, we propose a robust multi-dimensional evaluation protocol that integrates both objective metrics and LLM-based judgments, and empirically verify its high alignment with human experts. Our experiments reveal that while agentic pipelines outperform RAG baselines in fluency and structural quality, they still struggle with citation accuracy, highlighting key directions for future research.

Short-video platforms have rapidly become a new generation of information retrieval systems, where users formulate queries to access desired videos. However, user queries, especially long-tail ones, often suffer from spelling errors, incomplete phrasing, and ambiguous intent, resulting in mismatches between user expectations and retrieved results. While large language models (LLMs) have shown success in long-tail query rewriting within e-commerce, they struggle on short-video platforms, where proprietary content such as short videos, live streams, micro dramas, and user social networks falls outside their training distribution. To address this challenge, we introduce CardRewriter, an LLM-based framework that incorporates domain-specific knowledge to enhance long-tail query rewriting. For each query, our method aggregates multi-source knowledge relevant to the query and summarizes it into an informative and query-relevant knowledge card. This card then guides the LLM to better capture user intent and produce more effective query rewrites. We optimize CardRewriter using a two-stage training pipeline: supervised fine-tuning followed by group relative policy optimization, with a tailored reward system balancing query relevance and retrieval effectiveness. Offline experiments show that CardRewriter substantially improves rewriting quality for queries targeting proprietary content. Online A/B testing further confirms significant gains in long-view rate (LVR) and click-through rate (CTR), along with a notable reduction in initiative query reformulation rate (IQRR). Since September 2025, CardRewriter has been deployed on Kuaishou, one of China's largest short-video platforms, serving hundreds of millions of users daily.

Recent studies have shown that large language models (LLMs) can assess relevance and support information retrieval (IR) tasks such as document ranking and relevance judgment generation. However, the internal mechanisms by which off-the-shelf LLMs understand and operationalize relevance remain largely unexplored. In this article, we systematically investigate how different LLM modules contribute to relevance judgment through the lens of mechanistic interpretability. Using activation patching techniques, we analyze the roles of various model components and identify a multi-stage, progressive process in generating either pointwise or pairwise relevance judgment. Specifically, LLMs first extract query and document information in the early layers, then process relevance information according to instructions in the middle layers, and finally utilize specific attention heads in the later layers to generate relevance judgments in the required format. Our findings provide insights into the mechanisms underlying relevance assessment in LLMs, offering valuable implications for future research on leveraging LLMs for IR tasks.

Click modeling constitutes a pivotal area of study within information retrieval, as it provides insights into user search behavior and enables the extraction of valuable implicit relevance feedback from large-scale click logs. However, existing click models often rely on content-agnostic IDs to represent queries and documents. Additionally, they employ context-independent assumptions, such as the examination hypothesis, in modeling click probabilities. As a result, contemporary click models often fall short of capturing the influence of the diverse, multi-modal content found on modern Search Engine Result Pages (SERPs) and the intricate contextual interactions among heterogeneous search results. To address this issue, we propose a novel Dynamic Graph Neural Click Model (DGCM). The proposed model incorporates rich content and context information by jointly representing them as nodes in a dynamic graph and further leverages a dynamic graph attention network to predict users’ clicks at different timesteps. To demonstrate the effectiveness of the DGCM model, we conducted extensive experiments on two large-scale datasets with both content and click information: the public Sogou-SRR dataset and a proprietary dataset collected on the WeChat platform. The experimental results indicate that by capitalizing on the content and context information, DGCM outperforms existing click models in the click prediction and relevance estimation tasks.

This is a corrigendum for the article “A Content- and Context-Aware Click Model Based on Dynamic Graph Neural Networks” published in ACM Trans. Inf. Syst. 44, 3, Article 67 (March 2026), 27 pages.

Large language models (LLMs) have demonstrated remarkable capabilities across various professional domains, with their performance typically evaluated through standardized benchmarks. In the financial field, the stringent demands for professional accuracy and real-time data processing often necessitate the use of retrieval-augmented generation (RAG) techniques. However, the development of financial RAG benchmarks has been constrained by data confidentiality issues and the lack of dynamic data integration. To address this issue, we introduce FinS-Pilot, a novel benchmark for evaluating RAG systems in online financial applications. Constructed from real-world financial assistant interactions, our benchmark incorporates both real-time API data and text data, organized through an intent classification framework covering critical financial domains. The benchmark enables comprehensive evaluation of financial assistants' capabilities in handling both static knowledge and time-sensitive market information.Through systematic experiments with multiple Chinese leading LLMs, we demonstrate FinS-Pilot's effectiveness in identifying models suitable for financial applications while addressing the current gap in specialized evaluation tools for the financial domain. Our work contributes both a practical evaluation framework and a curated dataset to advance research in financial NLP systems. The code and dataset are accessible on GitHub.

To satisfy users' diverse information needs, the aggregated search systems need to integrate heterogeneous results, with rich but different structural information, from a variety of verticals, such as news search, video search, and product search. A key challenge in aggregated search is to effectively and efficiently retrieve the most relevant results among a large number of heterogeneous information from different verticals. With the development of deep learning and pre-trained language models (PLMs), many researchers resort to Dense Retrieval (DR) models for a unified, efficient embedding-based retrieval and a better retrieval performance. However, existing dense retrieval models have limitations in: 1) capturing the structural information of search results ; and 2) generalizing across different vertical domains where the search results have different or even unseen structures. In this paper, we aim to tackle these limitations, and propose an effective and efficient dense retrieval model for aggregated search. Specifically, we utilize a deep prompt-tuning technique to make the pre-training model easily applied to downstream vertical search tasks. To capture the structural knowledge, we design a Graph Neural Network (GNN)-based structure prompt, to prompt how text segments are organized in the vanilla semi-structured data. We further incorporate a distributional prompt to model the theme of each domain, and enhance cross-domain generalization. Extensive experiments on the real-world data collected from the WeChat Search demonstrate that for aggregated search tasks, our models can achieve better performance over existing retrieval models, and have the superior ability to generalize to the various or even unseen vertical search tasks.

Unbiased learning to rank (ULTR), which aims to learn unbiased ranking models from biased user behavior logs, plays an important role in Web search. Previous research on ULTR has studied a variety of biases in users' clicks, such as position bias, presentation bias, and outlier bias. However, existing work often assumes that the behavior logs are collected from an ''average'' user, neglecting the differences between different users in their search and browsing behaviors. In this paper, we introduce personalized factors into the ULTR framework, which we term the user-aware ULTR problem. Through a formal causal analysis of this problem, we demonstrate that existing user-oblivious methods are biased when different users have different preferences over queries and personalized propensities of examining documents. To address such a personalized bias, we propose a novel user-aware inverse-propensity-score estimator for learning-to-rank objectives. Specifically, our approach models the distribution of user browsing behaviors for each query and aggregates user-weighted examination probabilities to determine propensities. We theoretically prove that the user-aware estimator is unbiased under some mild assumptions and shows lower variance compared to the straightforward way of calculating a user-dependent propensity for each impression. Finally, we empirically verify the effectiveness of our user-aware estimator by conducting extensive experiments on two semi-synthetic datasets and a real-world dataset.

The widely adopted Cranfield paradigm fails to adequately capture user satisfaction due to a weak relevance-satisfaction correlation. Additionally, constructing test collections incurs high relevance annotation costs. To address these two limitations, we aim to explore the use of large language models (LLMs) to generate multilevel usefulness labels. We propose CLUE, a user-centric evaluation method that explicitly incorporates users' search context and behavior information into LLMs. Inspired by ordinal regression, it employs a cascade structure tailored for multilevel usefulness judgments. Our study shows that using CLUE, LLMs can effectively assess usefulness when provided with search context and behavior, outperforming third-party labeling methods. We also conduct ablation studies to explore the impact of each component in CLUE. Finally, we utilize the usefulness labels generated by CLUE to predict user satisfaction. Real-world experiments reveal that incorporating CLUE's usefulness labels significantly enhances the performance of the satisfaction prediction model.

Providing explanations to justify recommendations enhances user satisfaction and trust. Despite significant research on explanation generation methods, evaluating their quality remains a critical yet under-explored challenge. Although large language models (LLMs) have been used for automated evaluation of explanations, existing approaches fail to account for the highly personalized na- ture of explanation assessment, where user judgments towards the same explanations vary significantly. To address this, we pro- pose MAML+PEFT method that combines Model-Agnostic Meta- Learning (MAML) with LoRA-based parameter-efficient tuning to adapt LLMs for personalized explanation evaluation. Building on this, we introduce TSA-MAML (Task Similarity Aware MAML)+PEFT, which clusters users based on their estimated optimal model param- eters and learns group-specific meta models by leveraging implicit group distributions of user preferences. Experiments on synthetic and human-annotated datasets demonstrate superior alignment of MAML-based methods with human ratings in both generalization and few-shot adaptation settings. Additionally, we examine the cor- relation of MAML-based LLM-simulated human ratings with real online user behaviors on a large-scale recommendation platform, demonstrating the practical utility of our methods for real-world explainable recommendation systems.

Personalized recommender systems, which focus on predicting users' interests, have significantly enhanced user experiences across diverse applications. However, existing approaches implicitly model users' preferences through fitting the fine-grained labels (e.g., click labels), but often neglecting the coarse-grained interest information inherent in the inputs themselves. Relying solely on the fine-grained labels could bring negative impact on interest modeling and limit the performance, as the labels may carry inevitable noise in real-world scenarios. In addition, it is considerably demanding in terms of data for most existing approaches to effectively model users' multi-granularity interests with limited or no supporting examples, resulting in subpar performance due to the significant long-tail phenomenon. To tackle these issues, we propose a novel learning framework named the Multi-Granularity Interest Prediction Framework (MGIPF), for better modeling users' diverse interests. Unlike prior work, our key idea is to utilize both the coarse-grained and fine-grained interests for supervising the training of models. Specifically, we introduce a pseudo-labeling approach explicitly mining users' potential multi-granularity interests from the raw data, and propose coarse-grained interest prediction modules that collaborate to utilize the multi-granularity supervision signals to enhance the learning of low-frequency items. The corresponding coarse-grained losses are softly weighted, taking into account the varying confidence of potential multi-granularity preferences on positive and negative samples. Importantly, our framework is lightweight and adaptable, capable of being applied effectively to mainstream recommendation models, establishing a comprehensive end-to-end training process. Extensive experiments conducted on three publicly available datasets have demonstrated the efficacy of our approach. The code is available at https://github.com/GeWu-Lab/MGIPF.

User behavior and experience are important for improving information retrieval (IR) systems. While much research has focused on traditional IR systems, few studies have systematically examined user behavior and search experience with emerging generative IR systems. A key reason for this gap is the lack of publicly available toolkits to record user behavior and feedback in generative IR systems. We developed a comprehensive platform to collect user behavior and feedback on the generative IR system. This platform consists of: 1) a generative IR system that supports both API-based and customized retrieval-augmented generation (RAG) methods, 2) a user interface that logs various user behavior, including prompts, clicks, mouse movements, and scrolling, and 3) an annotation website that allows users to provide feedback. We believe the proposed platform has the potential to streamline data collection for user studies on generative IR systems, paving the way for future research on how users engage with and interact with these systems.

Unbiased learning to rank (ULTR), which utilizes historical click logs to train ranking models, has attracted much attention in the IR community. Previous studies on ULTR have focused on mitigating a variety of biases in click logs, such as position bias, trust bias, and presentation bias, to recover the true relevance of the query-document pairs. However, they overlooked the intrinsic distribution shifts between the training data and test data. In this paper, we first validate and analyze the distribution shift problem with a real-world ULTR dataset. To solve this problem, we propose distributionally robust unbiased learning to rank (DRO-ULTR) methods. Specifically, we design two kinds of group distributionally robust optimization (group-DRO) frameworks for the existing ULTR methods, one using the pointwise click prediction loss and the other using the listwise counterfactual ranking loss. Finally, we empirically verify the effectiveness of our DRO-ULTR methods by conducting extensive experiments on the real-world dataset.

Counterfactual learning to rank (CLTR) has attracted extensive attention in the IR community for its ability to leverage massive logged user interaction data to train ranking models. While the CLTR models can be theoretically unbiased when the user behavior assumption is correct and the propensity estimation is accurate, their effectiveness is usually empirically evaluated via simulation-based experiments due to a lack of widely available, large-scale, real click logs. However, many previous simulation-based experiments are somewhat limited because they may have one or more of the following deficiencies: 1) using a weak production ranker to generate initial ranked lists, 2) relying on a simplified user simulation model to simulate user clicks, and 3) generating a fixed number of synthetic click logs. As a result, the robustness of CLTR models in complex and diverse situations is largely unknown and needs further investigation.

Simulating user search behavior is a critical task in information retrieval, which can be employed for user behavior modeling, data augmentation, and system evaluation. Recent advancements in large language models (LLMs) have opened up new possibilities for generating human-like actions including querying, browsing, and clicking. In this work, we explore the integration of human-like thinking into search simulations by leveraging LLMs to simulate users' hidden cognitive processes. Specifically, given a search task and context, we prompt LLMs to first think like a human before executing the corresponding action. As existing search datasets do not include users' thought processes, we conducted a user study to collect a new dataset enriched with users' explicit thinking. We investigate the impact of incorporating such human-like thinking on simulation performance and apply supervised fine-tuning (SFT) to teach LLMs to emulate both human thinking and actions. Our experiments span two dimensions in leveraging LLMs for user simulation: (1) with or without explicit thinking, and (2) with or without fine-tuning on the thinking-augmented dataset. The results demonstrate the feasibility and potential of incorporating human-like thinking in user simulations, though performance improvements on some metrics remain modest. We believe this exploration provides new avenues and inspirations for advancing user behavior modeling in search simulations.

Large Language Models (LLMs) are increasingly employed in zero-shot documents ranking, yielding commendable results. However, several significant challenges still persist in LLMs for ranking: (1) LLMs are constrained by limited input length, precluding them from processing a large number of documents simultaneously; (2) The output document sequence is influenced by the input order of documents, resulting in inconsistent ranking outcomes; (3) Achieving a balance between cost and ranking performance is challenging. To tackle these issues, we introduce a novel documents ranking method called TourRank1. which is inspired by the sport tournaments, such as FIFA World Cup. Specifically, we 1) overcome the limitation in input length and reduce the ranking latency by incorporating a multi-stage grouping strategy similar to the parallel group stage of sport tournaments; 2) improve the ranking performance and robustness to input orders by using a points system to ensemble multiple ranking results. We test TourRank with different LLMs on the TREC DL datasets and the BEIR benchmark. The experimental results demonstrate that TourRank delivers state-of-the-art performance at a modest cost.

Search result diversification (SRD), which aims to ensure that documents in a ranking list cover a broad range of subtopics, is a significant and widely studied problem in Information Retrieval and Web Search. Existing methods primarily utilize a paradigm of ''greedy selection'', i.e., selecting one document with the highest diversity score at a time or optimize an approximation of the objective function. These approaches tend to be inefficient and are easily trapped in a suboptimal state. To address these challenges, we introduce Multi-Agent reinforcement learning (MARL) for search result DIVersity, which called MA4DIV. In this approach, each document is an agent and the search result diversification is modeled as a cooperative task among multiple agents. By modeling the SRD ranking problem as a cooperative MARL problem, this approach allows for directly optimizing the diversity metrics, such as α-NDCG, while achieving high training efficiency. We conducted experiments on public TREC datasets and a larger scale dataset in the industrial setting. The experiemnts show that MA4DIV achieves substantial improvements in both effectiveness and efficiency than existing baselines, especially on the industrial dataset.

Recent studies have demonstrated the effectiveness of using large language language models (LLMs) in passage ranking. The listwise approaches, such as RankGPT, have become new state-of-the-art in this task. However, the efficiency of RankGPT models is limited by the maximum context length and relatively high latency of LLM inference. To address these issues, in this paper, we propose PE-Rank, leveraging the single passage embedding as a good context compression for efficient listwise passage reranking. By treating each passage as a special token, we can directly input passage embeddings into LLMs, thereby reducing input length. Additionally, we introduce an inference method that dynamically constrains the decoding space to these special tokens, accelerating the decoding process. For adapting the model to reranking, we employ listwise learning to rank loss for training. Evaluation results on multiple benchmarks demonstrate that PE-Rank significantly improves efficiency in both prefilling and decoding, while maintaining competitive ranking effectiveness. The code is available at https://github.com/liuqi6777/pe_rank

In the information retrieval (IR) area, dense retrieval (DR) models use deep learning techniques to encode queries and passages into embedding space to compute their semantic relations. It is important for DR models to balance both efficiency and effectiveness. Pre-trained language models (PLMs), especially Transformer-based PLMs, have been proven to be effective encoders of DR models. However, the self-attention component in Transformer-based PLM results in a computational complexity that grows quadratically with sequence length, and thus exhibits a slow inference speed for long-text retrieval. Some recently proposed non-Transformer PLMs, especially the Mamba architecture PLMs, have demonstrated not only comparable effectiveness to Transformer-based PLMs on generative language tasks but also better efficiency due to linear time scaling in sequence length. This paper implements the Mamba Retriever to explore whether Mamba can serve as an effective and efficient encoder of DR model for IR tasks. We fine-tune the Mamba Retriever on the classic short-text MS MARCO passage ranking dataset and the long-text LoCoV0 dataset. Experimental results show that (1) on the MS MARCO passage ranking dataset and BEIR, the Mamba Retriever achieves comparable or better effectiveness compared to Transformer-based retrieval models, and the effectiveness grows with the size of the Mamba model; (2) on the long-text LoCoV0 dataset, the Mamba Retriever can extend to longer text length than its pre-trained length after fine-tuning on retrieval task, and it has comparable or better effectiveness compared to other long-text retrieval models; (3) the Mamba Retriever has superior inference speed for long-text retrieval. In conclusion, Mamba Retriever is both effective and efficient, making it a practical model, especially for long-text retrieval.

Providing natural language-based explanations to justify recommendations helps to improve users' satisfaction and gain users' trust. However, as current explanation generation methods are commonly trained with an objective to mimic existing user reviews, the generated explanations are often not aligned with the predicted ratings or some important features of the recommended items, and thus, are suboptimal in helping users make informed decision on the recommendation platform. To tackle this problem, we propose a flexible model-agnostic method named MMI (Maximizing Mutual Information) framework to enhance the alignment between the generated natural language explanations and the predicted rating/important item features. Specifically, we propose to use mutual information (MI) as a measure for the alignment and train a neural MI estimator. Then, we treat a well-trained explanation generation model as the backbone model and further fine-tune it through reinforcement learning with guidance from the MI estimator, which rewards a generated explanation that is more aligned with the predicted rating or a pre-defined feature of the recommended item. Experiments on three datasets demonstrate that our MMI framework can boost different backbone models, enabling them to outperform existing baselines in terms of alignment with predicted ratings and item features. Additionally, user studies verify that MI-enhanced explanations indeed facilitate users' decisions and are favorable compared with other baselines due to their better alignment properties.

Recently, geometric-inspired embedding methods draw research interests for their superior ability in representing transitive and asymmetric relations. A typical example, box embeddings, in which objects are parameterized as axis-aligned hyper-rectangles (i.e. boxes), can effectively model the partial orders and similarities between objects with the inclusion and overlapping relations of the boxes. However, the hard edges of the boxes present difficulties for gradient-based optimization. In this paper, we first identify two problems that may hinder the optimization of box embeddings, namely the zero-gradient problem and the vanishing gradient problem in high dimensional settings. Then, we propose a simple yet effective framework (Box++) to tackle these problems. For the first problem, Box++ combines the overlapping volumes and distance measures in optimizing box embeddings. The distance measures can naturally measure the “degree of disjointedness” for disjoint boxes and provide reasonable gradients for optimization. For the second problem, we theoretically prove that under certain conditions, the gradient would vanish exponentially, and therefore, make the optimization converges to suboptimal solutions. We further design a gradient normalization strategy for the Box++ models to alleviate this problem by rescaling the vanishing gradient. Extensive experiments on real-world datasets demonstrate the effectiveness of the proposed framework.

Scaling laws have been observed in a wide range of tasks, particularly in language generation. Previous studies have found that the performance of large language models adheres to predictable patterns with respect to the size of models and datasets. This helps us design training strategies effectively and efficiently, especially as large-scale training becomes increasingly resource-intensive. Yet, in dense retrieval, such scaling law has not been fully explored. In this study, we investigate how scaling affects the performance of dense retrieval models. We implement dense retrieval models with different numbers of parameters, and train them with various amounts of annotated data. We propose to use the contrastive entropy as the evaluation metric, which is continuous compared with discrete ranking metrics and thus can accurately reflect model performance. Results indicate that the performance of dense retrieval models follows a precise power-law scaling related to the model size and the number of annotations across different datasets and annotation methods. Additionally, we show that the scaling laws help optimize the training process, such as resolving the resource allocation problem under a budget constraint. We believe that these findings significantly contribute to understanding the scaling effect of dense retrieval models and offer meaningful guidance for future research.

The ability of the foundation models heavily relies on large-scale, diverse, and high-quality pretraining data. In order to improve data quality, researchers and practitioners often have to manually curate datasets from difference sources and develop dedicated data cleansing pipeline for each data repository. Lacking a unified data processing framework, this process is repetitive and cumbersome. To mitigate this issue, we propose a data processing framework that integrates a Processing Module which consists of a series of operators at different granularity levels, and an Analyzing Module which supports probing and evaluation of the refined data. The proposed framework is easy to use and highly flexible. In this demo paper, we first introduce how to use this framework with some example use cases and then demonstrate its effectiveness in improving the data quality with an automated evaluation with ChatGPT and an end-to-end evaluation in pretraining the GPT-2 model. The code and demonstration video are accessible on GitHub.

Due to the advantages in the cost-efficiency and reproducibility, user simulation has become a promising solution to the user-centric evaluation of information retrieval systems. Nonetheless, accurately simulating user search behaviors has long been a challenge, because users' actions in search are highly complex and driven by intricate cognitive processes such as learning, reasoning, and planning. Recently, Large Language Models (LLMs) have demonstrated remarked potential in simulating human-level intelligence and have been used in building autonomous agents for various tasks. However, the potential of using LLMs in simulating search behaviors has not yet been fully explored. In this paper, we introduce a LLM-based user search behavior simulator, USimAgent. The proposed simulator can simulate users' querying, clicking, and stopping behaviors during search, and thus, is capable of generating complete search sessions for specific search tasks. Empirical investigation on a real user behavior dataset shows that the proposed simulator outperforms existing methods in query generation and is comparable to traditional methods in predicting user clicks and stopping behaviors. These results not only validate the effectiveness of using LLMs for user simulation but also shed light on the development of a more robust and generic user simulators.

With the development of pre-trained language models, the dense retrieval models have become promising alternatives to the traditional retrieval models that rely on exact match and sparse bag-of-words representations. Different from most dense retrieval models using a bi-encoder to encode each query or document into a dense vector, the recently proposed late-interaction multi-vector models (i.e., ColBERT and COIL) achieve state-of-the-art retrieval effectiveness by using all token embeddings to represent documents and queries and modeling their relevance with a sum-of-max operation. However, these fine-grained representations may cause unacceptable storage overhead for practical search systems. In this study, we systematically analyze the matching mechanism of these late-interaction models and show that the sum-of-max operation heavily relies on the co-occurrence signals and some important words in the document. Based on these findings, we then propose several simple document pruning methods to reduce the storage overhead and compare the effectiveness of different pruning methods on different late-interaction models. We also leverage query pruning methods to further reduce the retrieval latency. We conduct extensive experiments on both in-domain and out-domain datasets and show that some of the used pruning methods can significantly improve the efficiency of these late-interaction models without substantially hurting their retrieval effectiveness.

Point-of-interest (POI) search is important for location-based services, such as navigation and online ride-hailing service. The goal of POI search is to find the most relevant destinations from a large-scale POI database given a text query. To improve the effectiveness and efficiency of POI search, most existing approaches are based on a multi-stage pipeline that consists of an efficiency-oriented retrieval stage and one or more effectiveness-oriented re-rank stages. In this article, we focus on the first efficiency-oriented retrieval stage of the POI search. We first identify the limitations of existing first-stage POI retrieval models in capturing the semantic-geography relationship and modeling the fine-grained geographical context information. Then, we propose a Geo-Enhanced Dense Retrieval framework for POI search to alleviate the above problems. Specifically, the proposed framework leverages the capacity of pre-trained language models (e.g., BERT) and designs a pre-training approach to better model the semantic match between the query prefix and POIs. With the POI collection, we first perform a token-level pre-training task based on a geographical-sensitive masked language prediction and design two retrieval-oriented pre-training tasks that link the address of each POI to its name and geo-location. With the user behavior logs collected from an online POI search system, we design two additional pre-training tasks based on users’ query reformulation behavior and the transitions between POIs. We also utilize a late-interaction network structure to model the fine-grained interactions between the text and geographical context information within an acceptable query latency. Extensive experiments on the real-world datasets collected from the Didichuxing application demonstrate that the proposed framework can achieve superior retrieval performance over existing first-stage POI retrieval methods.

Legal case retrieval is a special Information Retrieval (IR) task focusing on legal case documents. Depending on the downstream tasks of the retrieved case documents, users’ information needs in legal case retrieval could be significantly different from those in Web search and traditional ad hoc retrieval tasks. While there are several studies that retrieve legal cases based on text similarity, the underlying search intents of legal retrieval users, as shown in this article, are more complicated than that yet mostly unexplored. To this end, we present a novel hierarchical intent taxonomy of legal case retrieval. It consists of five intent types categorized by three criteria, i.e., search for Particular Case(s) , Characterization , Penalty , Procedure , and Interest . The taxonomy was constructed transparently and evaluated extensively through interviews, editorial user studies, and query log analysis. Through a laboratory user study, we reveal significant differences in user behavior and satisfaction under different search intents in legal case retrieval. Furthermore, we apply the proposed taxonomy to various downstream legal retrieval tasks, e.g., result ranking and satisfaction prediction, and demonstrate its effectiveness. Our work provides important insights into the understanding of user intents in legal case retrieval and potentially leads to better retrieval techniques in the legal domain, such as intent-aware ranking strategies and evaluation methodologies.

While dense retrieval has become a promising alternative to the traditional text retrieval models, such as BM25, some recent studies show that multi-vector dense retrieval models are more effective than the single-vector method in retrieval tasks. However, due to a lack of interpretability, why the multi-vector method outperforms its single-vector counterpart has not been fully studied. To fill this research gap, in this work, we investigate and compare the behaviors of single-vector and multi-vector models in retrieval. Specifically, we analyze the vocabulary distribution of dense representations by mapping them back to the sparse, vocabulary space. Our empirical findings show that the multi-vector representation has more lexical overlaps between queries and passages. Additionally, we show that this feature of multi-vector representation can enhance its ranking performance when a given passage can fulfill different information needs and thus can be retrieved by different queries. These results shed light on the internal mechanisms of multi-vector representation and may provide new perspectives for future research.

Recent studies have shown that Dense Retrieval (DR) techniques can significantly improve the performance of first-stage retrieval in IR systems. Despite its empirical effectiveness, the application of DR is still limited. In contrast to statistic retrieval models that rely on highly efficient inverted index solutions, DR models build dense embeddings that are difficult to be pre-processed with most existing search indexing systems. To avoid the expensive cost of brute-force search, the Approximate Nearest Neighbor (ANN) algorithm and corresponding indexes are widely applied to speed up the inference process of DR models. Unfortunately, while ANN can improve the efficiency of DR models, it usually comes with a significant price on retrieval performance.

Session search is a widely adopted technique in search engines that seeks to leverage the complete interaction history of a search session to better understand the information needs of users and provide more relevant ranking results. The vast majority of existing methods model a search session as a sequence of queries and previously clicked documents. However, if we simply represent a search session as a sequence we will lose the topological information in the original search session. It is non-trivial to model the intra-session interactions and complicated structural patterns among the previously issued queries, clicked documents, as well as the terms or entities that appeared in them. To solve this problem, in this paper, we propose a novel Session Search with Graph Classification Model (SSGC), which regards session search as a graph classification task on a heterogeneous graph that represents the search history in each session. To improve the performance of the graph classification, we design a specific pre-training strategy for our proposed GNN-based classification model. Extensive experiments on two public session search datasets demonstrate the effectiveness of our model in the session search task.

Reading is a vital and complex cognitive activity during users’ information-seeking process. Several studies have focused on understanding users’ reading behavior in desktop search. Their findings greatly contribute to the design of information retrieval models. However, little is known about how users read a result in mobile search, although search currently happens more frequently in mobile scenarios. In this paper, we conduct a lab-based user study to investigate users’ fine-grained reading behavior patterns in mobile search. We find that users’ reading attention allocation is strongly affected by several behavior biases, such as position and selection biases. Inspired by these findings, we propose a probabilistic generative model, the Passage-level Reading behavior Model (PRM), to model users’ reading behavior in mobile search. The PRM utilizes observable passage-level exposure and viewport duration events to infer users’ unobserved skimming event, reading event, and satisfaction perception during the reading process. Besides fitting the passage-level reading behavior, we utilize the fitted parameters of PRM to estimate the passage-level and document-level relevance. Experimental results show that PRM outperforms existing unsupervised relevance estimation models. PRM has strong interpretability and provides valuable insights into the understanding of how users seek and perceive useful information in mobile search.

Legal case retrieval, which aims to retrieve relevant cases given a query case, has drawn increasing research attention in recent years. While much research has worked on developing automatic retrieval models, how to characterize relevance in this specialized information retrieval (IR) task is still an open question. Towards an in-depth understanding of relevance judgments, we conduct a laboratory user study that involves 72 participants of different domain expertise. In the user study, we collect the relevance score along with detailed explanations for the relevance judgment and various measures of the judgment process. From the collected data, we observe that both the subjective (e.g., domain expertise) and objective (e.g., query/case property) factors influence the relevance judgment process. By investigating the collected user explanations, we identify task-specific patterns of user attention distribution and re-think the criteria for relevance judgments. Moreover, we investigate the similarity in attention distribution between models and users. Further, we propose a two-stage framework that utilizes user attention to improve relevance estimation for legal case retrieval. Our study sheds light on understanding relevance judgments in legal case retrieval and provides implications for improving the design of corresponding retrieval systems.

Result ranking is one of the major concerns for Web search technologies. Most existing methodologies rank search results in descending order of relevance. To model the interactions among search results, reinforcement learning (RL algorithms have been widely adopted for ranking tasks. However, the online training of RL methods is time and resource consuming at scale. As an alternative, learning ranking policies in the simulation environment is much more feasible and efficient. In this article, we propose two different simulation environments for the offline training of the RL ranking agent: the Context-aware Click Simulator (CCS) and the Fine-grained User Behavior Simulator with GAN (UserGAN). Based on the simulation environment, we also design a User Behavior Simulation for Reinforcement Learning (UBS4RL) re-ranking framework, which consists of three modules: a feature extractor for heterogeneous search results, a user simulator for collecting simulated user feedback, and a ranking agent for generation of optimized result lists. Extensive experiments on both simulated and practical Web search datasets show that (1) the proposed user simulators can capture and simulate fine-grained user behavior patterns by training on large-scale search logs, (2) the temporal information of user searching process is a strong signal for ranking evaluation, and (3) learning ranking policies from the simulation environment can effectively improve the search ranking performance.

The progress of recommender systems is hampered mainly by evaluation as it requires real-time interactions between humans and systems, which is too laborious and expensive. This issue is usually approached by utilizing the interaction history to conduct offline evaluation. However, existing datasets of user-item interactions are partially observed, leaving it unclear how and to what extent the missing interactions will influence the evaluation. To answer this question, we collect a fully-observed dataset from Kuaishou's online environment, where almost all 1,411 users have been exposed to all 3,327 items. To the best of our knowledge, this is the first real-world fully-observed data with millions of user-item interactions.

A retrieval model should not only interpolate the training data but also extrapolate well to the queries that are different from the training data. While neural retrieval models have demonstrated impressive performance on ad-hoc search benchmarks, we still know little about how they perform in terms of interpolation and extrapolation. In this paper, we demonstrate the importance of separately evaluating the two capabilities of neural retrieval models. Firstly, we examine existing ad-hoc search benchmarks from the two perspectives. We investigate the distribution of training and test data and find a considerable overlap in query entities, query intent, and relevance labels. This finding implies that the evaluation on these test sets is biased toward interpolation and cannot accurately reflect the extrapolation capacity. Secondly, we propose a novel evaluation protocol to separately evaluate the interpolation and extrapolation performance on existing benchmark datasets. It resamples the training and test data based on query similarity and utilizes the resampled dataset for training and evaluation. Finally, we leverage the proposed evaluation protocol to comprehensively revisit a number of widely-adopted neural retrieval models. Results show models perform differently when moving from interpolation to extrapolation. For example, representation-based retrieval models perform almost as well as interaction-based retrieval models in terms of interpolation but not extrapolation. Therefore, it is necessary to separately evaluate both interpolation and extrapolation performance and the proposed resampling method serves as a simple yet effective evaluation tool for future IR studies.

Recently, we have witnessed that deep learning-based approaches have been widely applied. Particularly, some applications involve data that are high dimensional, sparse or imbalanced, which are different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice and theory of deep learning on high-dimensional, sparse and imbalanced data.

Recently, pre-training methods tailored for IR tasks have achieved great success. However, as the mechanisms behind the performance improvement remain under-investigated, the interpretability and robustness of these pre-trained models still need to be improved. Axiomatic IR aims to identify a set of desirable properties expressed mathematically as formal constraints to guide the design of ranking models. Existing studies have already shown that considering certain axioms may help improve the effectiveness and interpretability of IR models. However, there still lack efforts of incorporating these IR axioms into pre-training methodologies. To shed light on this research question, we propose a novel pre-training method with \underlineA xiomatic \underlineRe gularization for ad hoc \underlineS earch (ARES). In the ARES framework, a number of existing IR axioms are re-organized to generate training samples to be fitted in the pre-training process. These training samples then guide neural rankers to learn the desirable ranking properties. Compared to existing pre-training approaches, ARES is more intuitive and explainable. Experimental results on multiple publicly available benchmark datasets have shown the effectiveness of ARES in both full-resource and low-resource (e.g., zero-shot and few-shot) settings. An intuitive case study also indicates that ARES has learned useful knowledge that existing pre-trained models (e.g., BERT and PROP) fail to possess. This work provides insights into improving the interpretability of pre-trained models and the guidance of incorporating IR axioms or human heuristics into pre-training methods.

Pre-trained language models (PLMs) have achieved great success in the area of Information Retrieval. Studies show that applying these models to ad-hoc document ranking can achieve better retrieval effectiveness. However, on the Web, most information is organized in the form of HTML web pages. In addition to the pure text content, the structure of the content organized by HTML tags is also an important part of the information delivered on a web page. Currently, such structured information is totally ignored by pre-trained models which are trained solely based on text content. In this paper, we propose to leverage large-scale web pages and their DOM (Document Object Model) tree structures to pre-train models for information retrieval. We argue that using the hierarchical structure contained in web pages, we can get richer contextual information for training better language models. To exploit this kind of information, we devise four pre-training objectives based on the structure of web pages, then pre-train a Transformer model towards these tasks jointly with traditional masked language model objective. Experimental results on two authoritative ad-hoc retrieval datasets prove that our model can significantly improve ranking performance compared to existing pre-trained models.

Proactive dialogue system is able to lead the conversation to a goal topic and has advantaged potential in bargain, persuasion, and negotiation. Current corpus-based learning manner limits its practical application in real-world scenarios. To this end, we contribute to advancing the study of the proactive dialogue policy to a more natural and challenging setting, i.e., interacting dynamically with users. Further, we call attention to the non-cooperative user behavior - the user talks about off-path topics when he/she is not satisfied with the previous topics introduced by the agent. We argue that the targets of reaching the goal topic quickly and maintaining a high user satisfaction are not always converged, because the topics close to the goal and the topics user preferred may not be the same. Towards this issue, we propose a new solution named I-Pro that can learn Proactive policy in the Interactive setting. Specifically, we learn the trade-off via a learned goal weight, which consists of four factors (dialogue turn, goal completion difficulty, user satisfaction estimation, and cooperative degree). The experimental results demonstrate I-Pro significantly outperforms baselines in terms of effectiveness and interpretability.

Asking clarifying questions is an interactive way to effectively clarify user intent. When a user submits a query, the search engine will return a clarifying question with several clickable items of sub-intents for clarification. According to the existing definition, the key to asking high-quality questions is to generate good descriptions for submitted queries and provided items. However, existing methods mainly based on static knowledge bases are difficult to find descriptions for many queries because of the lack of entities within these queries and their corresponding items. For such a query, it is unable to generate an informative question. To alleviate this problem, we propose leveraging top search results of the query to help generate better descriptions because we deem that the top retrieved documents contain rich and relevant contexts of the query. Specifically, we first design a rule-based algorithm to extract description candidates from search results and rank them by various human-designed features. Then, we apply an learning-to-rank model and another generative model for generalization and further improve the quality of clarifying questions. Experimental results show that our proposed methods can generate more readable and informative questions compared with existing methods. The results prove that search results can be utilized to improve users' search experience for search clarification in conversational search systems.

This paper presents a neural information retrieval pipeline that integrates cooperative learning of query reformulation and neural retrieval models. Our pipeline first exploits an automatic query reformulator to reformulate the user-issued query and then submits the reformulated query to the neural retrieval model. We simultaneously optimize the quality of reformulated queries and ranking performance with an alternate training strategy where query reformulator and neural retrieval model learn from the feedback of each other. Besides, we incorporate knowledge information into automatic query reformulation. The reformulated queries are further improved and contribute to a better ranking performance of the following neural retrieval model. We study two representative neural retrieval models KNRM and BERT in our pipeline. Experiments on two datasets show that our pipeline consistently improves the retrieval performance of the original neural retrieval models while only increases negligible time on automatic query reformulation.

Dense Retrieval (DR) has achieved state-of-the-art first-stage ranking effectiveness. However, the efficiency of most existing DR models is limited by the large memory cost of storing dense vectors and the time-consuming nearest neighbor search (NNS) in vector space. Therefore, we present RepCONC, a novel retrieval model that learns discrete Representations via CONstrained Clustering. RepCONC jointly trains dual-encoders and the Product Quantization (PQ) method to learn discrete document representations and enables fast approximate NNS with compact indexes. It models quantization as a constrained clustering process, which requires the document embeddings to be uniformly clustered around the quantization centroids and supports end-to-end optimization of the quantization method and dual-encoders. We theoretically demonstrate the importance of the uniform clustering constraint in RepCONC and derive an efficient approximate solution for constrained clustering by reducing it to an instance of the optimal transport problem. Besides constrained clustering, RepCONC further adopts a vector-based inverted file system (IVF) to support highly efficient vector search on CPUs. Extensive experiments on two popular ad-hoc retrieval benchmarks show that RepCONC achieves better ranking effectiveness than competitive vector quantization baselines under different compression ratio settings. It also substantially outperforms a wide range of existing retrieval models in terms of retrieval effectiveness, memory efficiency, and time efficiency.

Ranking has been one of the most important tasks in information retrieval. With the development of deep representation learning, many researchers propose to encode both the query and items into embedding vectors and rank the items according to the inner product or distance measures in the embedding space. However, the ranking models based on vector embeddings may have shortages in effectiveness and efficiency. For effectiveness, they lack the intrinsic ability to model the diversity and uncertainty of queries and items in ranking. For efficiency, nearest neighbor search in a large collection of item vectors can be costly. In this work, we propose to use the recently proposed probabilistic box embeddings for effective and efficient ranking, in which queries and items are parameterized as high-dimensional axis-aligned hyper-rectangles. For effectiveness, we utilize probabilistic box embeddings to model the diversity and uncertainty with the overlapping relations of the hyper-rectangles, and prove that such overlapping measure is a kernel function which can be adopted in other kernel-based methods. For efficiency, we propose a box embedding-based indexing method, which can safely filter irrelevant items and reduce the retrieval latency. We further design a training strategy to increase the proportion of irrelevant items that can be filtered by the index. Experiments on public datasets show that the box embeddings and the box embedding-based indexing approaches are effective and efficient in two ranking tasks: ad hoc retrieval and product recommendation.

Click models are widely used for user simulation, relevance inference, and evaluation in Web search. Most existing click models implicitly assume that users’ relevance judgment and behavior patterns are homogeneous. However, previous studies have shown that different users interact with search engines in rather different ways. Therefore, a unified click model can hardly capture the heterogeneity in users’ click behavior. To shed light on this research question, we propose a Click Model Personalization framework (CMP) that adaptively selects from global and local models for individual users. Different adaptive strategies are designed to personalize click behavior modeling only for specific users and queries. We also reveal that capturing personalized behavior patterns is more important than modeling personalized relevance assessments in constructing personalized click models. To evaluate the performance of the proposed CMP framework, we build a large-scale practical Personalized Web Search (PWS) dataset, which consists of the search logs of 1,249 users from a commercial search engine over six months. Experimental results show that the proposed CMP framework achieves significant performance improvements than the non-personalized click models in click prediction.

While batch evaluation plays a central part in Information Retrieval (IR) research, most evaluation metrics are based on user models which mainly focus on browsing and clicking behaviors. As users' perceived satisfaction may also be impacted by their search intent, constructing different user models across various search intent may help design better evaluation metrics. However, user intents are usually unobservable in practice. As query reformulating behaviors may reflect their search intents to a certain extent and highly correlate with users' perceived satisfaction for a specific query, these observable factors may be beneficial for the design of evaluation metrics. How to incorporate the search intent behind query reformulation into user behavior and satisfaction models remains under-investigated. To investigate the relationships among query reformulations, search intent, and user satisfaction, we explore a publicly available web search dataset and find that query reformulations can be a good proxy for inferring user intent, and therefore, reformulating actions may be beneficial for designing better web search effectiveness metrics. A group of Reformulation-Aware Metrics (RAMs) is then proposed to improve existing click model-based metrics. Experimental results on two public session datasets have shown that RAMs have significantly higher correlations with user satisfaction than existing evaluation metrics. In the robustness test, we have found that RAMs can achieve good performance when only a small proportion of satisfaction training labels are available. We further show that RAMs can be directly applied in a new dataset for offline evaluation once trained. This work shows the possibility of designing better evaluation metrics by incorporating fine-grained search context factors.

Relevance judgments play an essential role in the evaluation of information retrieval systems. As many different relevance judgment settings have been proposed in recent years, an evaluation metric to compare relevance judgments in different annotation settings has become a necessity. Traditional metrics, such as ĸ, Krippendorff's α and Φ have mainly focused on the inter-assessor consistency to evaluate the quality of relevance judgments. They encounter "reliable but useless" problem when employed to compare different annotation settings (e.g. binary judgment v.s. 4-grade judgment). Meanwhile, other existing popular metrics such as discriminative power (DP) are not designed to compare relevance judgments across different annotation settings, they therefore suffer from limitations, such as the requirement of result ranking lists from different systems. Therefore, how to design an evaluation metric to compare relevance judgments under different grade settings needs further investigation. In this work, we propose a novel metric named pairwise discriminative power (PDP) to evaluate the quality of relevance judgment collections. By leveraging a small amount of document-level preference tests, PDP estimates the discriminative ability of relevance judgments on separating ranking lists with various qualities. With comprehensive experiments on both synthetic and real-world datasets, we show that PDP maintains a high degree of consistency with annotation quality in various grade settings. Compared with existing metrics (e.g., Krippendorff's α, Φ, DP, etc), it provides reliable evaluation results with affordable additional annotation efforts.

Conversational search systems, such as Google Assistant and Microsoft Cortana, provide a new search paradigm where users are allowed, via natural language dialogues, to communicate with search systems. Evaluating such systems is very challenging since search results are presented in the format of natural language sentences. Given the unlimited number of possible responses, collecting relevance assessments for all the possible responses is infeasible. In this paper, we propose POSSCORE, a simple yet effective automatic evaluation method for conversational search. The proposed embedding-based metric takes the influence of part of speech (POS) of the terms in the response into account. To the best knowledge, our work is the first to systematically demonstrate the importance of incorporating syntactic information, such as POS labels, for conversational search evaluation. Experimental results demonstrate that our metrics can correlate with human preference, achieving significant improvements over state-of-the-art baseline metrics.

Recently, Information Retrieval community has witnessed fast-paced advances in Dense Retrieval (DR), which performs first-stage retrieval with embedding-based search. Despite the impressive ranking performance, previous studies usually adopt brute-force search to acquire candidates, which is prohibitive in practical Web search scenarios due to its tremendous memory usage and time cost. To overcome these problems, vector compression methods have been adopted in many practical embedding-based retrieval applications. One of the most popular methods is Product Quantization (PQ). However, although existing vector compression methods including PQ can help improve the efficiency of DR, they incur severely decayed retrieval performance due to the separation between encoding and compression. To tackle this problem, we present JPQ, which stands for Joint optimization of query encoding and Product Quantization. It trains the query encoder and PQ index jointly in an end-to-end manner based on three optimization strategies, namely ranking-oriented loss, PQ centroid optimization, and end-to-end negative sampling. We evaluate JPQ on two publicly available retrieval benchmarks. Experimental results show that JPQ significantly outperforms popular vector compression methods. Compared with previous DR models that use brute-force search, JPQ almost matches the best retrieval performance with 30x compression on index size. The compressed index further brings 10x speedup on CPU and 2x speedup on GPU in query latency.

Knowledge graphs are widely used in information retrieval as they can enhance our semantic understanding of queries and documents. The main idea is to consider entities and entity relationships as side information. Although existing work has achieved improvements in retrieval effectiveness by incorporating information from knowledge graphs into retrieval models, few studies have leveraged knowledge graphs in understanding users' search behavior. We investigate user behavior during session search from the perspective of a knowledge graph. We conduct a query log-based analysis of users' query reformulation and document clicking behavior. Based on a large-scale commercial query log and a knowledge graph, we find new user behavior patterns in terms of query reformulation and document clicking. Our study deepens our understanding of user behavior in session search and provides implications to help improve retrieval models with knowledge graphs.

Legal case retrieval is a specialized IR task aiming to retrieve supporting cases given a query case. While recent research efforts are committed to improving the automatic retrieval models' performances, little attention has been paid to the practical search interactions between users and systems in this task. Therefore, we focus on investigating user behavior in the scenario of legal case retrieval. Specifically, we conducted a laboratory user study that involved 45 participants majoring in law to collect users' rich interactions and relevance assessments. With the collected data, we first analyzed the characteristics of the search process in legal case retrieval practice. We observed significant differences between legal case retrieval and general web search in various search behavior. These differences highlight the necessity of in-depth investigating user behavior in legal case retrieval and re-thinking the application of related mechanisms developed based on the user models in Web search. Then we investigated factors that would influence search behavior from different perspectives, including task difficulty and domain expertise. Finally, we shed light on implicit feedback in legal case retrieval and designed a predictive model for relevance based on user behavior. Our work provides a better understanding of user interactions in the legal case retrieval process, which can benefit the design of the corresponding retrieval systems to support legal practitioners.

Ranking has always been one of the top concerns in information retrieval researches. For decades, the lexical matching signal has dominated the ad-hoc retrieval process, but solely using this signal in retrieval may cause the vocabulary mismatch problem. In recent years, with the development of representation learning techniques, many researchers turn to Dense Retrieval (DR) models for better ranking performance. Although several existing DR models have already obtained promising results, their performance improvement heavily relies on the sampling of training examples. Many effective sampling strategies are not efficient enough for practical usage, and for most of them, there still lacks theoretical analysis in how and why performance improvement happens. To shed light on these research questions, we theoretically investigate different training strategies for DR models and try to explain why hard negative sampling performs better than random sampling. Through the analysis, we also find that there are many potential risks in static hard negative sampling, which is employed by many existing training methods. Therefore, we propose two training strategies named a Stable Training Algorithm for dense Retrieval (STAR) and a query-side training Algorithm for Directly Optimizing Ranking pErformance (ADORE), respectively. STAR improves the stability of DR training process by introducing random negatives. ADORE replaces the widely-adopted static hard negative sampling method with a dynamic one to directly optimize the ranking performance. Experimental results on two publicly available retrieval benchmark datasets show that either strategy gains significant improvements over existing competitive baselines and a combination of them leads to the best performance.

Extracting valuable feedback information from user behavior logs is one of the major concerns in Web search studies. Among the tremendous efforts that aim to improve search performance with user behavior modeling, constructing click models is of vital importance because it provides a direct estimation of result relevance. Most existing click models assume that whether or not users click on results only depends on the examination probability and the content of the result. However, through a carefully designed user eye-tracking study, we found that users do not make click-through decisions in isolation. Instead, they also consider the context of a result (e.g., adjacent results). This finding leads to the design of a novel click model named Comparison-based Click Model (CBCM). Different from traditional examination hypotheses, CBCM introduces the concept of an examination viewport and assumes users click results after comparing adjacent results within the same viewport. The experimental results on a publicly available user behavior dataset demonstrate the effectiveness of CBCM. We also public our code of CBCM and dataset.

As queries submitted by users directly affect search experiences, how to organize queries has always been a research focus in Web search studies. While search request becomes complex and exploratory, many search sessions contain more than a single query thus reformulation becomes a necessity. To help users better formulate their queries in these complex search tasks, modern search engines usually provide a series of reformulation entries on search engine result pages (SERPs), i.e., query suggestions and related entities. However, few existing work have thoroughly studied why and how users perform query reformulations in these heterogeneous interfaces. Therefore, whether search engines provide sufficient assistance for users in reformulating queries remains under-investigated. To shed light on this research question, we conducted a field study to analyze fine-grained user reformulation behaviors including reformulation type, entry, reason, and the inspiration source with various search intents. Different from existing efforts that rely on external assessors to make judgments, in the field study we collect both implicit behavior signals and explicit user feedback information. Analysis results demonstrate that query reformulation behavior in Web search varies with the type of search tasks. We also found that the current query suggestion/related query recommendations provided by search engines do not offer enough help for users in complex search tasks. Based on the findings in our field study, we design a supervised learning framework to predict: 1) the reason behind each query reformulation, and 2) how users organize the reformulated query, both of which are novel challenges in this domain. This work provides insight into complex query reformulation behavior in Web search as well as the guidance for designing better query suggestion techniques in search engines.

Relevance measures the relation between query and document which contains several different dimensions, e.g., semantic similarity, topical relatedness, cognitive relevance (the relations in the aspect of knowledge), usefulness, timeliness, utility and so on. However, existing retrieval models mainly focus on semantic similarity and cognitive relevance while ignore other possible dimensions to model relevance. Topical relatedness, as an important dimension to measure relevance, is not well studied in existing neural information retrieval. In this paper, we propose a Topic Enhanced Knowledge-aware retrieval Model (TEKM) that jointly learns semantic similarity, knowledge relevance and topical relatedness to estimate relevance between query and document. We first construct a neural topic model to learn topical information and generate topic embeddings of a query. Then we combine the topic embeddings with a knowledge-aware retrieval model to estimate different dimensions of relevance. Specifically, we exploit kernel pooling to soft match topic embeddings with word and entity in a unified embedding space to generate fine-grained topical relatedness. The whole model is trained in an end-to-end manner. Experiments on a large-scale publicly available benchmark dataset show that TEKM outperforms existing retrieval models. Further analysis also shows how topic relatedness is modeled to improve traditional retrieval model with semantic similarity and knowledge relevance.

How to obtain an unbiased ranking model by learning to rank with biased user feedback is an important research question for IR. Existing work on unbiased learning to rank (ULTR) can be broadly categorized into two groups—the studies on unbiased learning algorithms with logged data, namely, the offline unbiased learning, and the studies on unbiased parameters estimation with real-time user interactions, namely, the online learning to rank. While their definitions of unbiasness are different, these two types of ULTR algorithms share the same goal—to find the best models that rank documents based on their intrinsic relevance or utility. However, most studies on offline and online unbiased learning to rank are carried in parallel without detailed comparisons on their background theories and empirical performance. In this article, we formalize the task of unbiased learning to rank and show that existing algorithms for offline unbiased learning and online learning to rank are just the two sides of the same coin. We evaluate eight state-of-the-art ULTR algorithms and find that many of them can be used in both offline settings and online environments with or without minor modifications. Further, we analyze how different offline and online learning paradigms would affect the theoretical foundation and empirical effectiveness of each algorithm on both synthetic and real search data. Our findings provide important insights and guidelines for choosing and deploying ULTR algorithms in practice.

Understanding user intent is essential for various retrieval tasks. By leveraging contextual information within sessions, e.g., query history and user click behaviors, search systems can capture user intent more accurately and thus perform better. However, most existing systems only consider intra-session contexts and may suffer from the problem of lacking contextual information, because short search sessions account for a large proportion in practical scenarios. We believe that in these scenarios, considering more contexts, e.g., cross-session dependencies, may help alleviate the problem and contribute to better performance. Therefore, we propose a novel Hybrid framework for Session Context Modeling (HSCM), which realizes session-level multi-task learning based on the self-attention mechanism. To alleviate the problem of lacking contextual information within current sessions, HSCM exploits the cross-session contexts by sampling user interactions under similar search intents in the historical sessions and further aggregating them into the local contexts. Besides, application of the self-attention mechanism rather than RNN-based frameworks in modeling session-level sequences also helps (1) better capture interactions within sessions, (2) represent the session contexts in parallelization. Experimental results on two practical search datasets show that HSCM not only outperforms strong baseline solutions such as HiNT, CARS, and BERTserini in document ranking, but also performs significantly better than most existing query suggestion methods. According to the results in an additional experiment, we have also found that HSCM is superior to most ranking models in click prediction.

Neural ranking models have recently gained much attention in Information Retrieval community and obtain good ranking performance. However, most of these retrieval models focus on capturing the textual matching signals between query and document but do not consider user behavior information that may be helpful for the retrieval task. Specifically, users' click and query reformulation behavior can be represented by a click-through bipartite graph and a session-flow graph, respectively. Such graph representations contain rich user behavior information and may help us better understand users' search intent beyond the textual information. In this study, we aim to incorporate this rich information encoded in these two graphs into existing neural ranking models.

Recent years have witnessed the success of deep neural networks in many research areas. The fundamental idea behind the design of most neural networks is to learn similarity patterns from data for prediction and inference, which lacks the ability of cognitive reasoning. However, the concrete ability of reasoning is critical to many theoretical and practical problems. On the other hand, traditional symbolic reasoning methods do well in making logical inference, but they are mostly hard rule-based reasoning, which limits their generalization ability to different tasks since difference tasks may require different rules. Both reasoning and generalization ability are important for prediction tasks such as recommender systems, where reasoning provides strong connection between user history and target items for accurate prediction, and generalization helps the model to draw a robust user portrait over noisy inputs.

A better understanding of users' reading behavior helps improve many information retrieval (IR) tasks, such as relevance estimation and document ranking. Existing research has already leveraged eye movement information to investigate user's reading process during document-level relevance judgments and the findings were adopted to build more effective ranking models. Recently, fine-grained (e.g., passage or sentence level) relevance judgments have been paid much attention to with the requirements in conversational search and QA systems. However, there is still a lack of thorough investigation on user's reading behavior during these kinds of interaction processes. To shed light on this research question, we investigate how users allocate their attention to passages of a document during the relevance judgment process. With the eye-tracking data collected in a laboratory study, we show that users pay more attention to the "key" passages which contain key useful information. Users tend to revisit these key passages several times to accumulate and verify the gathered information. With both content and user behavior features, we find that key passages can be predicted with supervised learning. We believe that this work contributes to better understanding users' reading behavior and may provide more explainability for relevance estimation.

Search applications such as image search, app search and product search are crucial parts of web search, which we denote as vertical search services. This tutorial will introduce the research and applications of user behavior modeling for vertical search. The bulk of the tutorial is devoted to covering research into behavior patterns, user behavior models and applications of user behavior data to refine evaluation metrics and ranking models for web-based vertical search.

Following the success of Cranfield-like evaluation approaches to evaluation in web search, web image search has also been evaluated with absolute judgments of (graded) relevance. However, recent research has found that collecting absolute relevance judgments may be difficult in image search scenarios due to the multi-dimensional nature of relevance for image results. Moreover, existing evaluation metrics based on absolute relevance judgments do not correlate well with search users' satisfaction perceptions in web image search.

Although BERT has shown its effectiveness in a number of IR-related tasks, especially document ranking, the understanding of its internal mechanism remains insufficient. To increase the explainability of the ranking process performed by BERT, we investigate a state-of-the-art BERT-based ranking model with focus on its attention mechanism and interaction behavior. Firstly, we look into the evolving of the attention distribution. It shows that in each step, BERT dumps redundant attention weights on tokens with high document frequency (such as periods). This may lead to a potential threat to the model robustness and should be considered in future studies. Secondly, we study how BERT models interactions between query and document and find that BERT aggregates document information to query token representations through their interactions, but extracts query-independent representations for document tokens. It indicates that it is possible to transform BERT into a more efficient representation-focused model. These findings help us better understand the ranking process by BERT and may inspire future improvement.

Recently session search evaluation has been paid more attention as a realistic search scenario usually involves multiple queries and interactions between users and systems. Evolved from model-based evaluation metrics for a single query, existing session-based metrics also follow a generic framework based on the cascade hypothesis. The cascade hypothesis assumes that lower-ranked search results and later-issued queries receive less attention from users and should therefore be assigned smaller weights when calculating evaluation metrics. This hypothesis gains much success in modeling search users' behavior and designing evaluation metrics, by explaining why users' attention decays on search engine result pages. However, recent studies have found that the recency effect also plays an important role in determining user satisfaction in search sessions. Especially, whether a user feels satisfied in the later-issued queries heavily influences his/her search satisfaction in the whole session. To take both the cascade hypothesis and the recency effect into the design of session search evaluation metrics, we propose Recency-aware Session-based Metrics (RSMs) to simultaneously characterize users' examination process with a browsing model and cognitive process with a utility accumulation model. With both self-constructed and public available user search behavior datasets, we show the effectiveness of proposed RSMs by comparing them with existing session-based metrics in the light of correlation with user satisfaction. We also find that the influence of the cascade and the recency effects varies dramatically among tasks with different difficulties and complexities, which suggests that we should use different model parameters for different types of search tasks. Our findings highlight the importance of investigating and utilizing cognitive effects besides examination hypotheses in search evaluation.

Evaluation metrics play an important role in the batch evaluation of IR systems. Based on a user model that describes how users interact with the rank list, an evaluation metric is defined to link the relevance scores of a list of documents to an estimation of system effectiveness and user satisfaction. Therefore, the validity of an evaluation metric has two facets: whether the underlying user model can accurately predict user behavior and whether the evaluation metric correlates well with user satisfaction. While a tremendous amount of work has been undertaken to design, evaluate, and compare different evaluation metrics, few studies have explored the consistency between these two facets of evaluation metrics. Specifically, we want to investigate whether the metrics that are well calibrated with user behavior data can perform as well in estimating user satisfaction. To shed light on this research question, we compare the performance of various metrics with the C/W/L Framework in estimating user satisfaction when they are optimized to fit observed user behavior. Experimental results on both self-collected and public available user search behavior datasets show that the metrics optimized to fit users' click behavior can perform as well as those calibrated with user satisfaction feedback. We also investigate the reliability in the calibration process of evaluation metrics to find out how much data is required for parameter tuning. Our findings provide empirical support for the consistency between user behavior modeling and satisfaction measurement, as well as guidance for tuning the parameters in evaluation metrics.

To better exploit the search logs, various click models have been proposed to extract implicit relevance feedback from user clicks. Most traditional click models are based on probability graphical models (PGMs) with manually designed dependencies. Recently, some researchers also adopt neural-based methods to improve the accuracy of click prediction. However, most of the existing click models only model user behavior in query level. As the previous iterations within the session may have an impact on the current search round, we can leverage these behavior signals to better model user behaviors. In this paper, we propose a novel neural- based Context-Aware Click Model (CACM) for Web search. CACM consists of a context-aware relevance estimator and an examination predictor. The relevance estimator utilizes session context infor- mation, i.e., the query sequence and clickthrough data, as well as the pre-trained embeddings learned from a session-flow graph to estimate the context-aware relevance of each search result. The examination predictor estimates the examination probability of each result. We further investigate several combination functions to integrate the context-aware relevance and examination probabil- ity into click prediction. Experiment results on a public Web search dataset show that CACM outperforms existing click models in both relevance estimation and click prediction tasks.

Examination is one of the most important user interactions in Web search. A number of works studied examination behavior in Web search and helped researchers better understand how users allocate their attention on search engine result pages (SERPs). Compared to desktop search, mobile search has a number of differences such as fewer results on the screen. These differences bring in mobile-specific factors affecting users' examination behavior. However, there still lacks research on users' attention allocation mechanism via viewports in mobile search. Therefore, we design a lab-based study to collect user's rich interaction behavior in mobile search. Based on the collected data, we first analyze how users examine SERPs and allocate their attention to heterogeneous results. Then we investigate the effect of mobile-specific factors and other common factors on users allocating attention. Finally, we apply the findings of user attention allocation from the user study into click model construction efforts, which significantly improves the state-of-the-art click model. Our work brings insights into a better understanding of users' interaction patterns in mobile search and may benefit other mobile search-related research.

Document ranking is one of the most studied but challenging problems in information retrieval (IR) research. A number of existing document ranking models capture relevance signals at the whole document level. Recently, more and more research has begun to address this problem from fine-grained document modeling. Several works leveraged fine-grained passage-level relevance signals in ranking models. However, most of these works focus on context-independent passage-level relevance signals and ignore the context information, which may lead to inaccurate estimation of passage-level relevance. In this paper, we investigate how information gain accumulates with passages when users sequentially read a document. We propose the context-aware Passage-level Cumulative Gain (PCG), which aggregates relevance scores of passages and avoids the need to formally split a document into independent passages. Next, we incorporate the patterns of PCG into a BERT-based sequential model called Passage-level Cumulative Gain Model (PCGM) to predict the PCG sequence. Finally, we apply PCGM to the document ranking task. Experimental results on two public ad hoc retrieval benchmark datasets show that PCGM outperforms most existing ranking models and also indicates the effectiveness of PCG signals. We believe that this work contributes to improving ranking performance and providing more explainability for document ranking.

Web search session data is precious for a wide range of Information Retrieval (IR) tasks, such as session search, query suggestion, click through rate (CTR) prediction and so on. Numerous studies have shown the great potential of considering context information for search system optimization. The well-known TREC Session Tracks have enhanced the development in this domain to a great extent. However, they are mainly collected via user studies or crowdsourcing experiments and normally contain only tens to thousands sessions, which are deficient for the investigation with more sophisticated models. To tackle this obstacle, we present a new dataset that contains 147,155 refined web search sessions with both click-based and human-annotated relevance labels. The sessions are sampled from a huge search log thus can reflect real search scenarios. The proposed dataset can support a wide range of session-level or task-based IR studies. As an example, we test several interactive search models with both the PSCM and human relevance labels provided by this dataset and report the performance as a reference for future studies of session search.

We investigated the learning process in search by conducting a log-based study involving registered job seekers of a commercial job search engine. The analysis shows that job search is a complex task: seekers usually submit multiple queries over sessions that can last days or even weeks. We find that querying, clicking, and job application rates change over time: job seekers tend to use more filters and a less diverse set of query terms. In terms of click and application behavior, we observed a significant decrease in click rate and query term diversity, as well as an increase in application rates. These trends are found to largely match information seeking models of learning in a complex search task. However, common behaviors are observed in the logs that suggest the existing models may not be sufficient to describe all of the users' learning and seeking processes.

In web image search, items users search for are images instead of Web pages or online services. Web image search constitutes a very important part of web search. Re-ranking is a trusted technique to improve retrieval effectiveness in web search. Previous work on re-ranking web image search results mainly focuses on intra-query information (e.g., human interactions with the initial list of the current query). Contextual information such as the query sequence and implicit user feedback provided during a search session prior to the current query is known to improve the performance of general web search but has so far not been used in web image search. The differences in result placement and interaction mechanisms of image search make the search process rather different from general Web search engines. Because of these differences, context-aware re-ranking models that have originally been developed for general web search cannot simply be applied to web image search. We propose CARM, a context-aware re-ranking model, a neural network-based framework to re-rank web image search results for a query based on previous interaction behavior in the search session in which the query was submitted. Specifically, we explore a hybrid encoder with an attention mechanism to model intra-query and inter-query user preferences for image results in a two-stage structure. We train context-aware re-ranking model (CARM) to jointly learn query and image representations so as to be able to deal with the multimodal characteristics of web image search. Extensive experiments are carried out on a commercial web image search dataset. The results show that CARM outperforms state-of-the-art baseline models in terms of personalized evaluation metrics. Also, CARM combines the original ranking can improve the original ranking on personalized ranking and relevance estimation. We make the implementation of CARM and relevant datasets publicly available to facilitate future studies.

Result ranking is one of the major concerns for Web search technologies. Most existing methodologies rank search results in descending order according to pointwise relevance estimation of single results. However, the dependency relationship between different search results are not taken into account. While search engine result pages contain more and more heterogenous components, a better ranking strategy should be a context-aware process and optimize result ranking globally. In this paper, we propose a novel framework which aims to improve context-aware listwise ranking performance by optimizing online evaluation metrics. The ranking problem is formalized as a Markov Decision Process (MDP) and solved with the reinforcement learning paradigm. To avoid the great cost to online systems during the training of the ranking model, we construct a virtual environment with millions of historical click logs to simulate the behavior of real users. Extensive experiments on both simulated and real datasets show that: 1) constructing a virtual environment can effectively leverage the large scale click logs and capture some important properties of real users. 2) the proposed framework can improve search ranking performance by a large margin.

User satisfaction is an important variable in Web search evaluation studies and has received more and more attention in recent years. Many studies regard user satisfaction as the ground truth for designing better evaluation metrics. However, most of the existing studies focus on designing Cranfield-like evaluation metrics to reflect user satisfaction at query-level. As information need becomes more and more complex, users often need multiple queries and multi-round search interactions to complete a search task (e.g. exploratory search). In those cases, how to characterize the user's satisfaction during a search session still remains to be investigated. In this paper, we collect a dataset through a laboratory study in which users need to complete some complex search tasks. With the help of hierarchical linear models (HLM), we try to reveal how user's query-level and session-level satisfaction are affected by different cognitive effects. A number of interesting findings are made. At query level, we found that although the relevance of top-ranked documents have important impacts (primacy effect), the average/maximum of perceived usefulness of clicked documents is a much better sign of user satisfaction. At session level, perceived satisfaction for a particular query is also affected by the other queries in the same session (anchor effect or expectation effect). We also found that session-level satisfaction correlates mostly with the last query in the session (recency effect). The findings will help us design better session-level user behavior models and corresponding evaluation metrics.

Retrieval models aim to estimate the relevance of a document to a certain query. Although existing retrieval models have gained much success in both deepening our understanding of information seeking behavior and constructing practical retrieval systems (e.g. Web search engines), we have to admit that the models work in a rather different manner than how humans make relevance judgments. In this paper, we aim to reexamine the existing models as well as to propose new ones based on the findings in how human read documents during relevance judgment. First, we summarize a number of reading heuristics from practical user behavior patterns, which are categorized into implicit and explicit heuristics. By reviewing a variety of existing retrieval models, we find that most of them only satisfy a part of these reading heuristics. To evaluate the effectiveness of each heuristic, we conduct an ablation study and find that most heuristics have positive impacts on retrieval performance. We further integrate all the effective heuristics into a new retrieval model named Reading Inspired Model (RIM). Specifically, implicit reading heuristics are incorporated into the model framework and explicit reading heuristics are modeled as a Markov Decision Process and learned by reinforcement learning. Experimental results on a large-scale public available benchmark dataset and two test sets from NTCIR WWW tasks show that RIM outperforms most existing models, which illustrates the effectiveness of the reading heuristics. We believe that this work contributes to constructing retrieval models with both higher retrieval performance and better explainability.

Compared to general web search, image search engines present results in a significantly different way, which leads to changes in user behavior patterns, and thus creates challenges for the existing evaluation mechanisms. In this paper, we pay attention to the context factor in the image search scenario. On the basis of a mean-variance analysis, we investigate the effects of context and find that evaluation metrics align with user satisfaction better when the returned image results have high variance. Furthermore, assuming that the image results a user has examined might affect her following judgments, we propose the Context-Aware Gain (CAG), a novel evaluation metric that incorporates the contextual effects within the well-known gain-discount framework. Our experiment results show that, with a proper combination of discount functions, the proposed context-aware evaluation metric can significantly improve the performances of offline metrics for image search evaluation, considering user satisfaction as the golden standard.

The understanding of the process of relevance judgment helps to inspire the design of retrieval models. Traditional retrieval models usually estimate relevance based on document-level signals. Recent works consider a more fine-grain, passage-level relevance information, which can further enhance retrieval performance. However, it lacks a detailed analysis of how passage-level relevance signals determine or influence the relevance judgment of the whole document. To investigate the role of passage-level relevance in the document-level relevance judgment, we construct an ad-hoc retrieval dataset with both passage-level and document-level relevance labels. A thorough analysis reveals that: 1) there is a strong correlation between the document-level relevance and the fractions of irrelevant passages to highly relevant passages; 2) the position, length and query similarity of passages play different roles in the determination of document-level relevance; 3) The sequential passage-level relevance within a document is a potential indicator for the document-level relevance. Based on the relationship between passage-level and document-level relevance, we also show that utilizing passage-level relevance signals can improve existing document ranking models. This study helps us better understand how users perceive relevance for a document and inspire the designing of novel ranking models leveraging fine-grain, passage-level relevance signals.

Explainable recommendation and search attempt to develop models or methods that not only generate high-quality recommendation or search results, but also intuitive explanations of the results for users or system designers, which can help to improve the system transparency, persuasiveness, trustworthiness, and effectiveness, etc. This is even more important in personalized search and recommendation scenarios, where users would like to know why a particular product, web page, news report, or friend suggestion exists in his or her own search and recommendation lists. The tutorial focuses on the research and application of explainable recommendation and search algorithms, as well as their application in real-world systems such as search engine, e-commerce and social networks. The tutorial aims at introducing and communicating explainable recommendation and search methods to the community, as well as gathering researchers and practitioners interested in this research direction for discussions, idea communications, and research promotions.

Machine Reading Comprehension (MRC) is one of the most challenging tasks in both NLP and IR researches. Recently, a number of deep neural models have been successfully adopted to some simplified MRC task settings, whose performances were close to or even better than human beings. However, these models still have large performance gaps with human beings in more practical settings, such as MS MARCO and DuReader datasets. Although there are many works studying human reading behavior, the behavior patterns in complex reading comprehension scenarios remain under-investigated. We believe that a better understanding of how human reads and allocates their attention during reading comprehension processes can help improve the performance of MRC tasks. In this paper, we conduct a lab study to investigate human's reading behavior patterns during reading comprehension tasks, where 32 users are recruited to take 60 distinct tasks. By analyzing the collected eye-tracking data and answers from participants, we propose a two-stage reading behavior model, in which the first stage is to search for possible answer candidates and the second stage is to generate the final answer through a comparison and verification process. We also find that human's attention distribution is affected by both question-dependent factors (e.g., answer and soft matching signal with questions) and question-independent factors (e.g., position, IDF and Part-of-Speech tags of words). We extract features derived from the two-stage reading behavior model to predict human's attention signals during reading comprehension, which significantly improves performance in the MRC task. Findings in our work may bring insight into the understanding of human reading and information seeking processes, and help the machine to better meet users' information needs.

Compared to general web search engines, web image search engines display results in a different way. In web image search, results are typically placed in a grid-based manner rather than a sequential result list. In this scenario, users can view results not only in a vertical direction but also in a horizontal direction. Moreover, pagination is usually not (explicitly) supported on image search search engine result pages (SERPs), and users can view results by scrolling down without having to click a “next page” button. These differences lead to different interaction mechanisms and user behavior patterns, which, in turn, create challenges to evaluation metrics that have originally been developed for general web search. While considerable effort has been invested in developing evaluation metrics for general web search, there has been relatively little effort to construct grid-based evaluation metrics.

Relevance estimation is among the most important tasks in the ranking of search results. Current methodologies mainly concentrate on text matching, link analysis, and user behavior models. However, users judge the relevance of search results directly from Search Engine Result Pages (SERPs), which provide valuable signals for reranking. In this article, we propose two different approaches to aggregate the visual, structure, as well as textual information sources of search results in relevance estimation. The first one is a late-fusion framework named Joint Relevance Estimation model (JRE). JRE estimates the relevance independently from screenshots, textual contents, and HTML source codes of search results and jointly makes the final decision through an inter-modality attention mechanism. The second one is an early-fusion framework named Tree-based Deep Neural Network (TreeNN), which embeds the texts and images into the HTML parse tree through a recursive process. To evaluate the performance of the proposed models, we construct a large-scale practical Search Result Relevance (SRR) dataset that consists of multiple information sources and relevance labels of over 60,000 search results. Experimental results show that the proposed two models achieve better performance than state-of-the-art ranking solutions as well as the original rankings of commercial search engines.

A series of click models has been proposed to extract accurate and unbiased relevance feedback from valuable yet noisy click-through data in search logs. Previous works have shown that users search behavior in mobile and desktop scenarios are rather different in many aspects, therefore, the click models designed for desktop search may not be effective in the mobile context. To address this problem, we propose two novel click models for mobile search: (1) Mobile Click Model (MCM), which models click necessity bias and examination satisfaction bias; (2) Viewport Time Click Model (VTCM), which further extends MCM by utilizing the viewport time. Extensive experiments on large-scale real mobile search logs show that: (1) MCM and VTCM outperform existing models in predicting users’ clicks and estimating result relevance; (2) MCM and VTCM can extract richer information, such as the click necessity of search results and the probability of user satisfaction, from mobile click logs; (3) By modeling the viewport time distributions of heterogeneous results, VTCM can bring a significant improvement over MCM in click prediction and relevance estimation tasks. Our proposed click models can help better understand user behavior patterns in mobile search and improve the ranking performance of mobile search engines.

With the rapid growth of mobile search, it is important to understand how users browse the mobile SERPs and allocate their limited attention to each result. To address this problem, we introduce a two-stage examination model that can separately capture the position bias with a skimming model and the attractiveness bias with an attractiveness model. The effectiveness of the proposed model is validated by using a dataset that contains explicit examination feedbacks from users. We further investigate user»s examination behaviors by analyzing the model parameters learned via EM algorithm. The results reveal some interesting findings such as how the skimming behavior is dependent on the previous examination sequence and what factors are associated with the attractiveness of search results on mobile SERPs.

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

Reading is a complex cognitive activity in many information retrieval related scenarios, such as relevance judgement and question answering. There exists plenty of works which model these processes as a matching problem, which focuses on how to estimate the relevance score between a document and a query. However, little is known about what happened during the reading process, i.e., how users allocate their attention while reading a document during a specific information retrieval task. We believe that a better understanding of this process can help us design better weighting functions inside the document and contributes to the improvement of ranking performance. In this paper, we focus on the reading process during relevance judgement task. We designed a lab-based user study to investigate human reading patterns in assessing a document, where users' eye movements and their labeled relevant text were collected, respectively. Through a systematic analysis into the collected data, we propose a two-stage reading model which consists of a preliminary relevance judgement stage (Stage 1) and a reading with preliminary relevance stage (Stage 2). In addition, we investigate how different behavior biases affect users' reading behaviors in these two stages. Taking these biases into consideration, we further build prediction models for user's reading attention. Experiment results show that query independent features outperform query dependent features, which indicates that users allocate attentions based on many signals other than query terms in this process. Our study sheds light on the understanding of users' attention allocation during relevance judgement and provides implications for improving the design of existing ranking models.

User satisfaction has been paid much attention to in recent Web search evaluation studies and regarded as the ground truth for designing better evaluation metrics. However, most existing studies are focused on the relationship between satisfaction and evaluation metrics at query-level. However, while search request becomes more and more complex, there are many scenarios in which multiple queries and multi-round search interactions are needed (e.g. exploratory search). In those cases, the relationship between session-level search satisfaction and session search evaluation metrics remain uninvestigated. In this paper, we analyze how users' perceptions of satisfaction accord with a series of session-level evaluation metrics. We conduct a laboratory study in which users are required to finish some complex search tasks and provide usefulness judgments of documents as well as session-level and query level satisfaction feedbacks. We test a number of popular session search evaluation metrics as well as different weighting functions. Experiment results show that query-level satisfaction is mainly decided by the clicked document that they think the most useful (maximum effect). While session-level satisfaction is highly correlated with the most recently issued queries (recency effect). We further propose a number of criteria for designing better session search evaluation metrics.

Users' click-through behavior is considered as a valuable yet noisy source of implicit relevance feedback for web search engines. A series of click models have therefore been proposed to extract accurate and unbiased relevance feedback from click logs. Previous works have shown that users' search behaviors in mobile and desktop scenarios are rather different in many aspects, therefore, the click models that were designed for desktop search may not be as effective in mobile context. To address this problem, we propose a novel Mobile Click Model (MCM) that models how users examine and click search results on mobile SERPs. Specifically, we incorporate two biases that are prevalent in mobile search into existing click models: 1) the click necessity bias that some results can bring utility and usefulness to users without being clicked; 2) the examination satisfaction bias that a user may feel satisfied and stop searching after examining a result with low click necessity. Extensive experiments on large-scale real mobile search logs show that: 1) MCM outperforms existing models in predicting users' click behavior in mobile search; 2) MCM can extract richer information, such as the click necessity of search results and the probability of user satisfaction, from mobile click logs. With this information, we can estimate the quality of different vertical results and improve the ranking of heterogeneous results in mobile search.

User interaction behavior is a valuable source of implicit relevance feedback. In Web image search a different type of search result presentation is used than in general Web search, which leads to different interaction mechanisms and user behavior. For example, image search results are self-contained, so that users do not need to click the results to view the landing page as in general Web search, which generates sparse click data. Also, two-dimensional result placement instead of a linear result list makes browsing behaviors more complex. Thus, it is hard to apply standard user behavior models (e.g., click models) developed for general Web search to Web image search. In this paper, we conduct a comprehensive image search user behavior analysis using data from a lab-based user study as well as data from a commercial search log. We then propose a novel interaction behavior model, called grid-based user browsing model (GUBM), whose design is motivated by observations from our data analysis. GUBM can both capture users' interaction behavior, including cursor hovering, and alleviate position bias. The advantages of GUBM are two-fold: (1) It is based on an unsupervised learning method and does not need manually annotated data for training. (2) It is based on user interaction features on search engine result pages (SERPs) and is easily transferable to other scenarios that have a grid-based interface such as video search engines. We conduct extensive experiments to test the performance of our model using a large-scale commercial image search log. Experimental results show that in terms of behavior prediction (perplexity), and topical relevance and image quality (normalized discounted cumulative gain (NDCG)), GUBM outperforms state-of-the-art baseline models as well as the original ranking. We make the implementation of GUBM and related datasets publicly available for future studies.

User satisfaction has been paid much attention to in recent Web search evaluation studies. Although satisfaction is often considered as an important symbol of search success, it doesn»t guarantee success in many cases, especially for complex search task scenarios. In this study, we investigate the differences between user satisfaction and search success, and try to adopt the findings to predict search success in complex search tasks. To achieve these research goals, we conduct a laboratory study in which search success and user satisfaction are annotated by domain expert assessors and search users, respectively. We find that both "Satisfaction with Failure" and "Unsatisfied Success" cases happen in these search tasks and together they account for as many as 40.3% of all search sessions. The factors (e.g. document readability and credibility) that lead to the inconsistency of search success and user satisfaction are also investigated and adopted to predict whether one search task is successful. Experimental results show that our proposed prediction method is effective in predicting search success.

People often conduct exploratory search to explore unfamiliar information space and learn new knowledge. While supporting the highly dynamic and interactive exploratory search is still challenging for the search system, we want to investigate which factors can make the exploratory search successful and satisfying from the user’s perspective. Previous research suggests that domain experts have different search strategies and are more successful in finding domain-specific information, but how the domain expertise level will influence users’ interaction and search outcomes in exploratory search, especially in different knowledge domains, is still unclear. In this work, via a carefully designed user study that involves 30 participants, we investigate the influence of domain expertise levels on the interaction and outcome of exploratory search in three different domains: environment, medicine, and politics. We record participants’ search behaviors, including their explicit feedback and eye fixation sequences, in a laboratory setting. With this dataset, we identify both domain-independent and domain-dependent effects on user behaviors and search outcomes. Our results extend existing research on the effect of domain expertise in search and suggest different strategies for exploiting domain expertise to support exploratory search in different knowledge domains.

Usefulness judgment measures the user-perceived amount of useful information for the search task in the current search context. Understanding and predicting usefulness judgment are crucial for developing user-centric evaluation methods and providing contextualize results according to the search context. With a dataset collected in a laboratory user study, we systematically investigate the effects of a variety of content, context, and behavior factors on usefulness judgments and find that while user behavior factors are most important in determining usefulness judgments, content and context factors also have significant effects on it. We further adopt these factors as features to build prediction models for usefulness judgments. An AUC score of 0.909 in binary usefulness classification and a Pearson's correlation coefficient of 0.694 in usefulness regression demonstrate the effectiveness of our models. Our study sheds light on the understanding of the dynamics of the user-perceived usefulness of documents in a search session and provides implications for the evaluation and design of Web search engines.

Relevance is a fundamental concept in information retrieval (IR) studies. It is however often observed that relevance as annotated by secondary assessors may not necessarily mean usefulness and satisfaction perceived by users. In this study, we confirm the difference by a laboratory study in which we collect relevance annotations by external assessors, usefulness and user satisfaction information by users, for a set of search tasks. We also find that a measure based on usefulness rather than relevance annotated has a better correlation with user satisfaction. However, we show that external assessors are capable of annotating usefulness when provided with more search context information. In addition, we also show that it is possible to generate automatically usefulness labels when some training data is available. Our findings explain why traditional system-centric evaluation metrics are not well aligned with user satisfaction and suggest that a usefulness-based evaluation method can be defined to better reflect the quality of search systems perceived by the users.