Nick Craswell

Paul Thomas, Seth Spielman, Nick Craswell et al. · microsoft-research

Relevance labels, which indicate whether a search result is valuable to a searcher, are key to evaluating and optimising search systems. The best way to capture the true preferences of users is to ask them for their careful feedback on which results would be useful, but this approach does not scale to produce a large number of labels. Getting relevance labels at scale is usually done with third-party labellers, who judge on behalf of the user, but there is a risk of low-quality data if the labeller doesn't understand user needs. To improve quality, one standard approach is to study real users through interviews, user studies and direct feedback, find areas where labels are systematically disagreeing with users, then educate labellers about user needs through judging guidelines, training and monitoring. This paper introduces an alternate approach for improving label quality. It takes careful feedback from real users, which by definition is the highest-quality first-party gold data that can be derived, and develops an large language model prompt that agrees with that data. We present ideas and observations from deploying language models for large-scale relevance labelling at Bing, and illustrate with data from TREC. We have found large language models can be effective, with accuracy as good as human labellers and similar capability to pick the hardest queries, best runs, and best groups. Systematic changes to the prompts make a difference in accuracy, but so too do simple paraphrases. To measure agreement with real searchers needs high-quality "gold" labels, but with these we find that models produce better labels than third-party workers, for a fraction of the cost, and these labels let us train notably better rankers.

Sebastian Hofstätter, Nick Craswell, Bhaskar Mitra et al. · microsoft-research

Recently, several dense retrieval (DR) models have demonstrated competitive performance to term-based retrieval that are ubiquitous in search systems. In contrast to term-based matching, DR projects queries and documents into a dense vector space and retrieves results via (approximate) nearest neighbor search. Deploying a new system, such as DR, inevitably involves tradeoffs in aspects of its performance. Established retrieval systems running at scale are usually well understood in terms of effectiveness and costs, such as query latency, indexing throughput, or storage requirements. In this work, we propose a framework with a set of criteria that go beyond simple effectiveness measures to thoroughly compare two retrieval systems with the explicit goal of assessing the readiness of one system to replace the other. This includes careful tradeoff considerations between effectiveness and various cost factors. Furthermore, we describe guardrail criteria, since even a system that is better on average may have systematic failures on a minority of queries. The guardrails check for failures on certain query characteristics and novel failure types that are only possible in dense retrieval systems. We demonstrate our decision framework on a Web ranking scenario. In that scenario, state-of-the-art DR models have surprisingly strong results, not only on average performance but passing an extensive set of guardrail tests, showing robustness on different query characteristics, lexical matching, generalization, and number of regressions. It is impossible to predict whether DR will become ubiquitous in the future, but one way this is possible is through repeated applications of decision processes such as the one presented here.

Zhenduo Wang, Yuancheng Tu, Corby Rosset et al. · tsinghua

A long-standing challenge for search and conversational assistants is query intention detection in ambiguous queries. Asking clarifying questions in conversational search has been widely studied and considered an effective solution to resolve query ambiguity. Existing work have explored various approaches for clarifying question ranking and generation. However, due to the lack of real conversational search data, they have to use artificial datasets for training, which limits their generalizability to real-world search scenarios. As a result, the industry has shown reluctance to implement them in reality, further suspending the availability of real conversational search interaction data. The above dilemma can be formulated as a cold start problem of clarifying question generation and conversational search in general. Furthermore, even if we do have large-scale conversational logs, it is not realistic to gather training data that can comprehensively cover all possible queries and topics in open-domain search scenarios. The risk of fitting bias when training a clarifying question retrieval/generation model on incomprehensive dataset is thus another important challenge. In this work, we innovatively explore generating clarifying questions in a zero-shot setting to overcome the cold start problem and we propose a constrained clarifying question generation system which uses both question templates and query facets to guide the effective and precise question generation. The experiment results show that our method outperforms existing state-of-the-art zero-shot baselines by a large margin. Human annotations to our model outputs also indicate our method generates 25.2\% more natural questions, 18.1\% more useful questions, 6.1\% less unnatural and 4\% less useless questions.

Shivani Upadhyay, Nandan Thakur, Ronak Pradeep et al.

The second edition of the TREC Retrieval Augmented Generation (RAG) Track advances research on systems that integrate retrieval and generation to address complex, real-world information needs. Building on the foundation of the inaugural 2024 track, this year's challenge introduces long, multi-sentence narrative queries to better reflect the deep search task with the growing demand for reasoning-driven responses. Participants are tasked with designing pipelines that combine retrieval and generation while ensuring transparency and factual grounding. The track leverages the MS MARCO V2.1 corpus and employs a multi-layered evaluation framework encompassing relevance assessment, response completeness, attribution verification, and agreement analysis. By emphasizing multi-faceted narratives and attribution-rich answers from over 150 submissions this year, the TREC 2025 RAG Track aims to foster innovation in creating trustworthy, context-aware systems for retrieval augmented generation.

Shivani Upadhyay, Ronak Pradeep, Nandan Thakur et al.

The application of large language models to provide relevance assessments presents exciting opportunities to advance information retrieval, natural language processing, and beyond, but to date many unknowns remain. This paper reports on the results of a large-scale evaluation (the TREC 2024 RAG Track) where four different relevance assessment approaches were deployed in situ: the "standard" fully manual process that NIST has implemented for decades and three different alternatives that take advantage of LLMs to different extents using the open-source UMBRELA tool. This setup allows us to correlate system rankings induced by the different approaches to characterize tradeoffs between cost and quality. We find that in terms of nDCG@20, nDCG@100, and Recall@100, system rankings induced by automatically generated relevance assessments from UMBRELA correlate highly with those induced by fully manual assessments across a diverse set of 77 runs from 19 teams. Our results suggest that automatically generated UMBRELA judgments can replace fully manual judgments to accurately capture run-level effectiveness. Surprisingly, we find that LLM assistance does not appear to increase correlation with fully manual assessments, suggesting that costs associated with human-in-the-loop processes do not bring obvious tangible benefits. Overall, human assessors appear to be stricter than UMBRELA in applying relevance criteria. Our work validates the use of LLMs in academic TREC-style evaluations and provides the foundation for future studies.

Hossein A. Rahmani, Varsha Ramineni, Emine Yilmaz et al.

Test collections are crucial for evaluating Information Retrieval (IR) systems. Creating a diverse set of user queries for these collections can be challenging, and obtaining relevance judgments, which indicate how well retrieved documents match a query, is often costly and resource-intensive. Recently, generating synthetic datasets using Large Language Models (LLMs) has gained attention in various applications. While previous work has used LLMs to generate synthetic queries or documents to improve ranking models, using LLMs to create synthetic test collections is still relatively unexplored. Previous work~\cite{rahmani2024synthetic} showed that synthetic test collections have the potential to be used for system evaluation, however, more analysis is needed to validate this claim. In this paper, we thoroughly investigate the reliability of synthetic test collections constructed using LLMs, where LLMs are used to generate synthetic queries, labels, or both. In particular, we examine the potential biases that might occur when such test collections are used for evaluation. We first empirically show the presence of such bias in evaluation results and analyse the effects it might have on system evaluation. We further validate the presence of such bias using a linear mixed-effects model. Our analysis shows that while the effect of bias present in evaluation results obtained using synthetic test collections could be significant, for e.g.~computing absolute system performance, its effect may not be as significant in comparing relative system performance. Codes and data are available at: https://github.com/rahmanidashti/BiasSyntheticData.

Ronak Pradeep, Nandan Thakur, Sahel Sharifymoghaddam et al.

Did you try out the new Bing Search? Or maybe you fiddled around with Google AI~Overviews? These might sound familiar because the modern-day search stack has recently evolved to include retrieval-augmented generation (RAG) systems. They allow searching and incorporating real-time data into large language models (LLMs) to provide a well-informed, attributed, concise summary in contrast to the traditional search paradigm that relies on displaying a ranked list of documents. Therefore, given these recent advancements, it is crucial to have an arena to build, test, visualize, and systematically evaluate RAG-based search systems. With this in mind, we propose the TREC 2024 RAG Track to foster innovation in evaluating RAG systems. In our work, we lay out the steps we've made towards making this track a reality -- we describe the details of our reusable framework, Ragnarök, explain the curation of the new MS MARCO V2.1 collection choice, release the development topics for the track, and standardize the I/O definitions which assist the end user. Next, using Ragnarök, we identify and provide key industrial baselines such as OpenAI's GPT-4o or Cohere's Command R+. Further, we introduce a web-based user interface for an interactive arena allowing benchmarking pairwise RAG systems by crowdsourcing. We open-source our Ragnarök framework and baselines to achieve a unified standard for future RAG systems.

Hamed Zamani, Nick Craswell

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

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al. · microsoft-research

This is the third year of the TREC Deep Learning track. As in previous years, we leverage the MS MARCO datasets that made hundreds of thousands of human annotated training labels available for both passage and document ranking tasks. In addition, this year we refreshed both the document and the passage collections which also led to a nearly four times increase in the document collection size and nearly $16$ times increase in the size of the passage collection. Deep neural ranking models that employ large scale pretraininig continued to outperform traditional retrieval methods this year. We also found that single stage retrieval can achieve good performance on both tasks although they still do not perform at par with multistage retrieval pipelines. Finally, the increase in the collection size and the general data refresh raised some questions about completeness of NIST judgments and the quality of the training labels that were mapped to the new collections from the old ones which we discuss in this report.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al. · microsoft-research

This is the fourth year of the TREC Deep Learning track. As in previous years, we leverage the MS MARCO datasets that made hundreds of thousands of human annotated training labels available for both passage and document ranking tasks. In addition, this year we also leverage both the refreshed passage and document collections that were released last year leading to a nearly $16$ times increase in the size of the passage collection and nearly four times increase in the document collection size. Unlike previous years, in 2022 we mainly focused on constructing a more complete test collection for the passage retrieval task, which has been the primary focus of the track. The document ranking task was kept as a secondary task, where document-level labels were inferred from the passage-level labels. Our analysis shows that similar to previous years, deep neural ranking models that employ large scale pretraining continued to outperform traditional retrieval methods. Due to the focusing our judging resources on passage judging, we are more confident in the quality of this year's queries and judgments, with respect to our ability to distinguish between runs and reuse the dataset in future. We also see some surprises in overall outcomes. Some top-performing runs did not do dense retrieval. Runs that did single-stage dense retrieval were not as competitive this year as they were last year.

Hossein A. Rahmani, Nick Craswell, Emine Yilmaz et al.

Test collections play a vital role in evaluation of information retrieval (IR) systems. Obtaining a diverse set of user queries for test collection construction can be challenging, and acquiring relevance judgments, which indicate the appropriateness of retrieved documents to a query, is often costly and resource-intensive. Generating synthetic datasets using Large Language Models (LLMs) has recently gained significant attention in various applications. In IR, while previous work exploited the capabilities of LLMs to generate synthetic queries or documents to augment training data and improve the performance of ranking models, using LLMs for constructing synthetic test collections is relatively unexplored. Previous studies demonstrate that LLMs have the potential to generate synthetic relevance judgments for use in the evaluation of IR systems. In this paper, we comprehensively investigate whether it is possible to use LLMs to construct fully synthetic test collections by generating not only synthetic judgments but also synthetic queries. In particular, we analyse whether it is possible to construct reliable synthetic test collections and the potential risks of bias such test collections may exhibit towards LLM-based models. Our experiments indicate that using LLMs it is possible to construct synthetic test collections that can reliably be used for retrieval evaluation.

Ronak Pradeep, Nandan Thakur, Shivani Upadhyay et al.

This report provides an initial look at partial results from the TREC 2024 Retrieval-Augmented Generation (RAG) Track. We have identified RAG evaluation as a barrier to continued progress in information access (and more broadly, natural language processing and artificial intelligence), and it is our hope that we can contribute to tackling the many challenges in this space. The central hypothesis we explore in this work is that the nugget evaluation methodology, originally developed for the TREC Question Answering Track in 2003, provides a solid foundation for evaluating RAG systems. As such, our efforts have focused on "refactoring" this methodology, specifically applying large language models to both automatically create nuggets and to automatically assign nuggets to system answers. We call this the AutoNuggetizer framework. Within the TREC setup, we are able to calibrate our fully automatic process against a manual process whereby nuggets are created by human assessors semi-manually and then assigned manually to system answers. Based on initial results across 21 topics from 45 runs, we observe a strong correlation between scores derived from a fully automatic nugget evaluation and a (mostly) manual nugget evaluation by human assessors. This suggests that our fully automatic evaluation process can be used to guide future iterations of RAG systems.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

This is the fifth year of the TREC Deep Learning track. As in previous years, we leverage the MS MARCO datasets that made hundreds of thousands of human-annotated training labels available for both passage and document ranking tasks. We mostly repeated last year's design, to get another matching test set, based on the larger, cleaner, less-biased v2 passage and document set, with passage ranking as primary and document ranking as a secondary task (using labels inferred from passage). As we did last year, we sample from MS MARCO queries that were completely held out, unused in corpus construction, unlike the test queries in the first three years. This approach yields a more difficult test with more headroom for improvement. Alongside the usual MS MARCO (human) queries from MS MARCO, this year we generated synthetic queries using a fine-tuned T5 model and using a GPT-4 prompt. The new headline result this year is that runs using Large Language Model (LLM) prompting in some way outperformed runs that use the "nnlm" approach, which was the best approach in the previous four years. Since this is the last year of the track, future iterations of prompt-based ranking can happen in other tracks. Human relevance assessments were applied to all query types, not just human MS MARCO queries. Evaluation using synthetic queries gave similar results to human queries, with system ordering agreement of $τ=0.8487$. However, human effort was needed to select a subset of the synthetic queries that were usable. We did not see clear evidence of bias, where runs using GPT-4 were favored when evaluated using synthetic GPT-4 queries, or where runs using T5 were favored when evaluated on synthetic T5 queries.

Nandan Thakur, Ronak Pradeep, Shivani Upadhyay et al.

Retrieval-augmented generation (RAG) enables large language models (LLMs) to generate answers with citations from source documents containing "ground truth", thereby reducing system hallucinations. A crucial factor in RAG evaluation is "support", whether the information in the cited documents supports the answer. To this end, we conducted a large-scale comparative study of 45 participant submissions on 36 topics to the TREC 2024 RAG Track, comparing an automatic LLM judge (GPT-4o) against human judges for support assessment. We considered two conditions: (1) fully manual assessments from scratch and (2) manual assessments with post-editing of LLM predictions. Our results indicate that for 56% of the manual from-scratch assessments, human and GPT-4o predictions match perfectly (on a three-level scale), increasing to 72% in the manual with post-editing condition. Furthermore, by carefully analyzing the disagreements in an unbiased study, we found that an independent human judge correlates better with GPT-4o than a human judge, suggesting that LLM judges can be a reliable alternative for support assessment. To conclude, we provide a qualitative analysis of human and GPT-4o errors to help guide future iterations of support assessment.

Ronak Pradeep, Nandan Thakur, Shivani Upadhyay et al.

Large Language Models (LLMs) have significantly enhanced the capabilities of information access systems, especially with retrieval-augmented generation (RAG). Nevertheless, the evaluation of RAG systems remains a barrier to continued progress, a challenge we tackle in this work by proposing an automatic evaluation framework that is validated against human annotations. We believe that the nugget evaluation methodology provides a solid foundation for evaluating RAG systems. This approach, originally developed for the TREC Question Answering (QA) Track in 2003, evaluates systems based on atomic facts that should be present in good answers. Our efforts focus on "refactoring" this methodology, where we describe the AutoNuggetizer framework that specifically applies LLMs to both automatically create nuggets and automatically assign nuggets to system answers. In the context of the TREC 2024 RAG Track, we calibrate a fully automatic approach against strategies where nuggets are created manually or semi-manually by human assessors and then assigned manually to system answers. Based on results from a community-wide evaluation, we observe strong agreement at the run level between scores derived from fully automatic nugget evaluation and human-based variants. The agreement is stronger when individual framework components such as nugget assignment are automated independently. This suggests that our evaluation framework provides tradeoffs between effort and quality that can be used to guide the development of future RAG systems. However, further research is necessary to refine our approach, particularly in establishing robust per-topic agreement to diagnose system failures effectively.

Miriam Wanner, Leif Azzopardi, Paul Thomas et al.

Existing methods for evaluating the factuality of large language model (LLM) responses treat all claims as equally important. This results in misleading evaluations when vital information is missing or incorrect as it receives the same weight as peripheral details, raising the question: how can we reliably detect such differences when there are errors in key information? Current approaches that measure factuality tend to be insensitive to omitted or false key information. To investigate this lack of sensitivity, we construct VITALERRORS, a benchmark of 6,733 queries with minimally altered LLM responses designed to omit or falsify key information. Using this dataset, we demonstrate the insensitivities of existing evaluation metrics to key information errors. To address this gap, we introduce VITAL, a set of metrics that provide greater sensitivity in measuring the factuality of responses by incorporating the relevance and importance of claims with respect to the query. Our analysis demonstrates that VITAL metrics more reliably detect errors in key information than previous methods. Our dataset, metrics, and analysis provide a foundation for more accurate and robust assessment of LLM factuality.

Gabriella Kazai, Bhaskar Mitra, Anlei Dong et al.

Traditional information retrieval (IR) ranking models process the full text of documents. Newer models based on Transformers, however, would incur a high computational cost when processing long texts, so typically use only snippets from the document instead. The model's input based on a document's URL, title, and snippet (UTS) is akin to the summaries that appear on a search engine results page (SERP) to help searchers decide which result to click. This raises questions about when such summaries are sufficient for relevance estimation by the ranking model or the human assessor, and whether humans and machines benefit from the document's full text in similar ways. To answer these questions, we study human and neural model based relevance assessments on 12k query-documents sampled from Bing's search logs. We compare changes in the relevance assessments when only the document summaries and when the full text is also exposed to assessors, studying a range of query and document properties, e.g., query type, snippet length. Our findings show that the full text is beneficial for humans and a BERT model for similar query and document types, e.g., tail, long queries. A closer look, however, reveals that humans and machines respond to the additional input in very different ways. Adding the full text can also hurt the ranker's performance, e.g., for navigational queries.

Jianfeng Gao, Chenyan Xiong, Paul Bennett et al.

A conversational information retrieval (CIR) system is an information retrieval (IR) system with a conversational interface which allows users to interact with the system to seek information via multi-turn conversations of natural language, in spoken or written form. Recent progress in deep learning has brought tremendous improvements in natural language processing (NLP) and conversational AI, leading to a plethora of commercial conversational services that allow naturally spoken and typed interaction, increasing the need for more human-centric interactions in IR. As a result, we have witnessed a resurgent interest in developing modern CIR systems in both research communities and industry. This book surveys recent advances in CIR, focusing on neural approaches that have been developed in the last few years. This book is based on the authors' tutorial at SIGIR'2020 (Gao et al., 2020b), with IR and NLP communities as the primary target audience. However, audiences with other background, such as machine learning and human-computer interaction, will also find it an accessible introduction to CIR. We hope that this book will prove a valuable resource for students, researchers, and software developers. This manuscript is a working draft. Comments are welcome.

Sebastian Hofstätter, Bhaskar Mitra, Hamed Zamani et al.

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

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

Evaluation efforts such as TREC, CLEF, NTCIR and FIRE, alongside public leaderboard such as MS MARCO, are intended to encourage research and track our progress, addressing big questions in our field. However, the goal is not simply to identify which run is "best", achieving the top score. The goal is to move the field forward by developing new robust techniques, that work in many different settings, and are adopted in research and practice. This paper uses the MS MARCO and TREC Deep Learning Track as our case study, comparing it to the case of TREC ad hoc ranking in the 1990s. We show how the design of the evaluation effort can encourage or discourage certain outcomes, and raising questions about internal and external validity of results. We provide some analysis of certain pitfalls, and a statement of best practices for avoiding such pitfalls. We summarize the progress of the effort so far, and describe our desired end state of "robust usefulness", along with steps that might be required to get us there.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

The TREC Deep Learning (DL) Track studies ad hoc search in the large data regime, meaning that a large set of human-labeled training data is available. Results so far indicate that the best models with large data may be deep neural networks. This paper supports the reuse of the TREC DL test collections in three ways. First we describe the data sets in detail, documenting clearly and in one place some details that are otherwise scattered in track guidelines, overview papers and in our associated MS MARCO leaderboard pages. We intend this description to make it easy for newcomers to use the TREC DL data. Second, because there is some risk of iteration and selection bias when reusing a data set, we describe the best practices for writing a paper using TREC DL data, without overfitting. We provide some illustrative analysis. Finally we address a number of issues around the TREC DL data, including an analysis of reusability.

Bhaskar Mitra, Sebastian Hofstatter, Hamed Zamani et al.

The Transformer-Kernel (TK) model has demonstrated strong reranking performance on the TREC Deep Learning benchmark -- and can be considered to be an efficient (but slightly less effective) alternative to other Transformer-based architectures that employ (i) large-scale pretraining (high training cost), (ii) joint encoding of query and document (high inference cost), and (iii) larger number of Transformer layers (both high training and high inference costs). Since, a variant of the TK model -- called TKL -- has been developed that incorporates local self-attention to efficiently process longer input sequences in the context of document ranking. In this work, we propose a novel Conformer layer as an alternative approach to scale TK to longer input sequences. Furthermore, we incorporate query term independence and explicit term matching to extend the model to the full retrieval setting. We benchmark our models under the strictly blind evaluation setting of the TREC 2020 Deep Learning track and find that our proposed architecture changes lead to improved retrieval quality over TKL. Our best model also outperforms all non-neural runs ("trad") and two-thirds of the pretrained Transformer-based runs ("nnlm") on NDCG@10.

Jimmy Lin, Daniel Campos, Nick Craswell et al.

Leaderboards are a ubiquitous part of modern research in applied machine learning. By design, they sort entries into some linear order, where the top-scoring entry is recognized as the "state of the art" (SOTA). Due to the rapid progress being made in information retrieval today, particularly with neural models, the top entry in a leaderboard is replaced with some regularity. These are touted as improvements in the state of the art. Such pronouncements, however, are almost never qualified with significance testing. In the context of the MS MARCO document ranking leaderboard, we pose a specific question: How do we know if a run is significantly better than the current SOTA? We ask this question against the backdrop of recent IR debates on scale types: in particular, whether commonly used significance tests are even mathematically permissible. Recognizing these potential pitfalls in evaluation methodology, our study proposes an evaluation framework that explicitly treats certain outcomes as distinct and avoids aggregating them into a single-point metric. Empirical analysis of SOTA runs from the MS MARCO document ranking leaderboard reveals insights about how one run can be "significantly better" than another that are obscured by the current official evaluation metric (MRR@100).

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

This is the second year of the TREC Deep Learning Track, with the goal of studying ad hoc ranking in the large training data regime. We again have a document retrieval task and a passage retrieval task, each with hundreds of thousands of human-labeled training queries. We evaluate using single-shot TREC-style evaluation, to give us a picture of which ranking methods work best when large data is available, with much more comprehensive relevance labeling on the small number of test queries. This year we have further evidence that rankers with BERT-style pretraining outperform other rankers in the large data regime.

Bhaskar Mitra, Sebastian Hofstatter, Hamed Zamani et al.

We benchmark Conformer-Kernel models under the strict blind evaluation setting of the TREC 2020 Deep Learning track. In particular, we study the impact of incorporating: (i) Explicit term matching to complement matching based on learned representations (i.e., the "Duet principle"), (ii) query term independence (i.e., the "QTI assumption") to scale the model to the full retrieval setting, and (iii) the ORCAS click data as an additional document description field. We find evidence which supports that all three aforementioned strategies can lead to improved retrieval quality.

Bhaskar Mitra, Sebastian Hofstatter, Hamed Zamani et al.

The Transformer-Kernel (TK) model has demonstrated strong reranking performance on the TREC Deep Learning benchmark---and can be considered to be an efficient (but slightly less effective) alternative to BERT-based ranking models. In this work, we extend the TK architecture to the full retrieval setting by incorporating the query term independence assumption. Furthermore, to reduce the memory complexity of the Transformer layers with respect to the input sequence length, we propose a new Conformer layer. We show that the Conformer's GPU memory requirement scales linearly with input sequence length, making it a more viable option when ranking long documents. Finally, we demonstrate that incorporating explicit term matching signal into the model can be particularly useful in the full retrieval setting. We present preliminary results from our work in this paper.

Bodo Billerbeck, Justin Zobel, Nicholas Lester et al.

Search techniques make use of elementary information such as term frequencies and document lengths in computation of similarity weighting. They can also exploit richer statistics, in particular the number of documents in which any two terms co-occur. In this paper we propose alternative methods for computing this statistic, a challenging task because the number of distinct pairs of terms is vast -- around 100,000 in a typical 1000-word news article, for example. In contrast, we do not employ approximation algorithms, as we want to be able to find exact counts. We explore their efficiency, finding that a naïve approach based on a dictionary is indeed very slow, while methods based on a combination of inverted indexes and linear scanning provide both massive speed-ups and better observed asymptotic behaviour. Our careful implementation shows that, with our novel list-pairs approach it is possible to process over several hundred thousand documents per hour.

Hamed Zamani, Gord Lueck, Everest Chen et al.

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

Nick Craswell, Daniel Campos, Bhaskar Mitra et al.

Users of Web search engines reveal their information needs through queries and clicks, making click logs a useful asset for information retrieval. However, click logs have not been publicly released for academic use, because they can be too revealing of personally or commercially sensitive information. This paper describes a click data release related to the TREC Deep Learning Track document corpus. After aggregation and filtering, including a k-anonymity requirement, we find 1.4 million of the TREC DL URLs have 18 million connections to 10 million distinct queries. Our dataset of these queries and connections to TREC documents is of similar size to proprietary datasets used in previous papers on query mining and ranking. We perform some preliminary experiments using the click data to augment the TREC DL training data, offering by comparison: 28x more queries, with 49x more connections to 4.4x more URLs in the corpus. We present a description of the dataset's generation process, characteristics, use in ranking and suggest other potential uses.

Hamed Zamani, Bhaskar Mitra, Everest Chen et al.

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

Sebastian Hofstätter, Hamed Zamani, Bhaskar Mitra et al.

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

Emine Yilmaz, Nick Craswell, Bhaskar Mitra et al.

As deep learning based models are increasingly being used for information retrieval (IR), a major challenge is to ensure the availability of test collections for measuring their quality. Test collections are generated based on pooling results of various retrieval systems, but until recently this did not include deep learning systems. This raises a major challenge for reusable evaluation: Since deep learning based models use external resources (e.g. word embeddings) and advanced representations as opposed to traditional methods that are mainly based on lexical similarity, they may return different types of relevant document that were not identified in the original pooling. If so, test collections constructed using traditional methods are likely to lead to biased and unfair evaluation results for deep learning (neural) systems. This paper uses simulated pooling to test the fairness and reusability of test collections, showing that pooling based on traditional systems only can lead to biased evaluation of deep learning systems.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

The Deep Learning Track is a new track for TREC 2019, with the goal of studying ad hoc ranking in a large data regime. It is the first track with large human-labeled training sets, introducing two sets corresponding to two tasks, each with rigorous TREC-style blind evaluation and reusable test sets. The document retrieval task has a corpus of 3.2 million documents with 367 thousand training queries, for which we generate a reusable test set of 43 queries. The passage retrieval task has a corpus of 8.8 million passages with 503 thousand training queries, for which we generate a reusable test set of 43 queries. This year 15 groups submitted a total of 75 runs, using various combinations of deep learning, transfer learning and traditional IR ranking methods. Deep learning runs significantly outperformed traditional IR runs. Possible explanations for this result are that we introduced large training data and we included deep models trained on such data in our judging pools, whereas some past studies did not have such training data or pooling.

Bhaskar Mitra, Nick Craswell

This report discusses three submissions based on the Duet architecture to the Deep Learning track at TREC 2019. For the document retrieval task, we adapt the Duet model to ingest a "multiple field" view of documents---we refer to the new architecture as Duet with Multiple Fields (DuetMF). A second submission combines the DuetMF model with other neural and traditional relevance estimators in a learning-to-rank framework and achieves improved performance over the DuetMF baseline. For the passage retrieval task, we submit a single run based on an ensemble of eight Duet models.

Hongfei Zhang, Xia Song, Chenyan Xiong et al.

This paper presents GEneric iNtent Encoder (GEN Encoder) which learns a distributed representation space for user intent in search. Leveraging large scale user clicks from Bing search logs as weak supervision of user intent, GEN Encoder learns to map queries with shared clicks into similar embeddings end-to-end and then finetunes on multiple paraphrase tasks. Experimental results on an intrinsic evaluation task - query intent similarity modeling - demonstrate GEN Encoder's robust and significant advantages over previous representation methods. Ablation studies reveal the crucial role of learning from implicit user feedback in representing user intent and the contributions of multi-task learning in representation generality. We also demonstrate that GEN Encoder alleviates the sparsity of tail search traffic and cuts down half of the unseen queries by using an efficient approximate nearest neighbor search to effectively identify previous queries with the same search intent. Finally, we demonstrate distances between GEN encodings reflect certain information seeking behaviors in search sessions.

Bhaskar Mitra, Corby Rosset, David Hawking et al.

Classical information retrieval (IR) methods, such as query likelihood and BM25, score documents independently w.r.t. each query term, and then accumulate the scores. Assuming query term independence allows precomputing term-document scores using these models---which can be combined with specialized data structures, such as inverted index, for efficient retrieval. Deep neural IR models, in contrast, compare the whole query to the document and are, therefore, typically employed only for late stage re-ranking. We incorporate query term independence assumption into three state-of-the-art neural IR models: BERT, Duet, and CKNRM---and evaluate their performance on a passage ranking task. Surprisingly, we observe no significant loss in result quality for Duet and CKNRM---and a small degradation in the case of BERT. However, by operating on each query term independently, these otherwise computationally intensive models become amenable to offline precomputation---dramatically reducing the cost of query evaluations employing state-of-the-art neural ranking models. This strategy makes it practical to use deep models for retrieval from large collections---and not restrict their usage to late stage re-ranking.

Corby Rosset, Bhaskar Mitra, Chenyan Xiong et al.

Axiomatic information retrieval (IR) seeks a set of principle properties desirable in IR models. These properties when formally expressed provide guidance in the search for better relevance estimation functions. Neural ranking models typically contain a large number of parameters. The training of these models involve a search for appropriate parameter values based on large quantities of labeled examples. Intuitively, axioms that can guide the search for better traditional IR models should also help in better parameter estimation for machine learning based rankers. This work explores the use of IR axioms to augment the direct supervision from labeled data for training neural ranking models. We modify the documents in our dataset along the lines of well-known axioms during training and add a regularization loss based on the agreement between the ranking model and the axioms on which version of the document---the original or the perturbed---should be preferred. Our experiments show that the neural ranking model achieves faster convergence and better generalization with axiomatic regularization.

Bhaskar Mitra, Nick Craswell

We propose several small modifications to Duet---a deep neural ranking model---and evaluate the updated model on the MS MARCO passage ranking task. We report significant improvements from the proposed changes based on an ablation study.

Hamed Zamani, Bhaskar Mitra, Xia Song et al.

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

Bhaskar Mitra, Nick Craswell

Neural ranking models for information retrieval (IR) use shallow or deep neural networks to rank search results in response to a query. Traditional learning to rank models employ machine learning techniques over hand-crafted IR features. By contrast, neural models learn representations of language from raw text that can bridge the gap between query and document vocabulary. Unlike classical IR models, these new machine learning based approaches are data-hungry, requiring large scale training data before they can be deployed. This tutorial introduces basic concepts and intuitions behind neural IR models, and places them in the context of traditional retrieval models. We begin by introducing fundamental concepts of IR and different neural and non-neural approaches to learning vector representations of text. We then review shallow neural IR methods that employ pre-trained neural term embeddings without learning the IR task end-to-end. We introduce deep neural networks next, discussing popular deep architectures. Finally, we review the current DNN models for information retrieval. We conclude with a discussion on potential future directions for neural IR.

Bhaskar Mitra, Fernando Diaz, Nick Craswell

In recent years, the information retrieval (IR) community has witnessed the first successful applications of deep neural network models to short-text matching and ad-hoc retrieval. It is exciting to see the research on deep neural networks and IR converge on these tasks of shared interest. However, the two communities have less in common when it comes to the choice of programming languages. Indri, an indexing framework popularly used by the IR community, is written in C++, while Torch, a popular machine learning library for deep learning, is written in the light-weight scripting language Lua. To bridge this gap, we introduce Luandri (pronounced "laundry"), a simple interface for exposing the search capabilities of Indri to Torch models implemented in Lua.

Payal Bajaj, Daniel Campos, Nick Craswell et al.

We introduce a large scale MAchine Reading COmprehension dataset, which we name MS MARCO. The dataset comprises of 1,010,916 anonymized questions---sampled from Bing's search query logs---each with a human generated answer and 182,669 completely human rewritten generated answers. In addition, the dataset contains 8,841,823 passages---extracted from 3,563,535 web documents retrieved by Bing---that provide the information necessary for curating the natural language answers. A question in the MS MARCO dataset may have multiple answers or no answers at all. Using this dataset, we propose three different tasks with varying levels of difficulty: (i) predict if a question is answerable given a set of context passages, and extract and synthesize the answer as a human would (ii) generate a well-formed answer (if possible) based on the context passages that can be understood with the question and passage context, and finally (iii) rank a set of retrieved passages given a question. The size of the dataset and the fact that the questions are derived from real user search queries distinguishes MS MARCO from other well-known publicly available datasets for machine reading comprehension and question-answering. We believe that the scale and the real-world nature of this dataset makes it attractive for benchmarking machine reading comprehension and question-answering models.

Bhaskar Mitra, Fernando Diaz, Nick Craswell

Models such as latent semantic analysis and those based on neural embeddings learn distributed representations of text, and match the query against the document in the latent semantic space. In traditional information retrieval models, on the other hand, terms have discrete or local representations, and the relevance of a document is determined by the exact matches of query terms in the body text. We hypothesize that matching with distributed representations complements matching with traditional local representations, and that a combination of the two is favorable. We propose a novel document ranking model composed of two separate deep neural networks, one that matches the query and the document using a local representation, and another that matches the query and the document using learned distributed representations. The two networks are jointly trained as part of a single neural network. We show that this combination or `duet' performs significantly better than either neural network individually on a Web page ranking task, and also significantly outperforms traditional baselines and other recently proposed models based on neural networks.

Fernando Diaz, Bhaskar Mitra, Nick Craswell

Continuous space word embeddings have received a great deal of attention in the natural language processing and machine learning communities for their ability to model term similarity and other relationships. We study the use of term relatedness in the context of query expansion for ad hoc information retrieval. We demonstrate that word embeddings such as word2vec and GloVe, when trained globally, underperform corpus and query specific embeddings for retrieval tasks. These results suggest that other tasks benefiting from global embeddings may also benefit from local embeddings.

Bhaskar Mitra, Eric Nalisnick, Nick Craswell et al.

A fundamental goal of search engines is to identify, given a query, documents that have relevant text. This is intrinsically difficult because the query and the document may use different vocabulary, or the document may contain query words without being relevant. We investigate neural word embeddings as a source of evidence in document ranking. We train a word2vec embedding model on a large unlabelled query corpus, but in contrast to how the model is commonly used, we retain both the input and the output projections, allowing us to leverage both the embedding spaces to derive richer distributional relationships. During ranking we map the query words into the input space and the document words into the output space, and compute a query-document relevance score by aggregating the cosine similarities across all the query-document word pairs. We postulate that the proposed Dual Embedding Space Model (DESM) captures evidence on whether a document is about a query term in addition to what is modelled by traditional term-frequency based approaches. Our experiments show that the DESM can re-rank top documents returned by a commercial Web search engine, like Bing, better than a term-matching based signal like TF-IDF. However, when ranking a larger set of candidate documents, we find the embeddings-based approach is prone to false positives, retrieving documents that are only loosely related to the query. We demonstrate that this problem can be solved effectively by ranking based on a linear mixture of the DESM and the word counting features.