We wish to measure the information coverage of an ad hoc retrieval algorithm, that is, how much of the range of available relevant information is covered by the search results. Information coverage is a central aspect for retrieval, especially when the retrieval system is integrated with generative models in a retrieval-augmented generation (RAG) system. The classic metrics for ad hoc retrieval, precision and recall, reward a system as more relevant documents are retrieved. However, since relevance in ad hoc test collections is defined for a document without any relation to other documents that might contain the same information, high recall is sufficient but not necessary to ensure coverage. The same is true for other metrics such as rank-biased precision (RBP), normalized discounted cumulative gain (nDCG), and mean average precision (MAP). Test collections developed around the notion of diversity ranking in web search incorporate multiple aspects that support a concept of coverage in the web domain. In this work, we construct a benchmark, CoverageBench, for evaluating information coverage made from existing collections. This suite offers researchers a unified testbed spanning multiple genres and tasks. All topics, nuggets, relevance labels, and baseline rankings are released on Hugging Face Datasets, along with instructions for accessing the publicly available document collections.

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.

Test collections are information retrieval tools that allow researchers to quickly and easily evaluate ranking algorithms. While test collections have become an integral part of IR research, the process of data creation involves significant manual annotation effort, which often makes it very expensive and time-consuming. Consequently, test collections could become too small when the budget is limited, which may lead to unstable evaluations. As a cheaper alternative, recent studies have proposed the use of large language models (LLMs) to completely replace human assessors. However, while LLMs seem to somewhat correlate with human judgments, their predictions are not perfect and often show bias. Thus, a complete replacement with LLMs is argued to be too risky and not fully reliable.

Retrieval-augmented generation (RAG) enables large language models (LLMs) to generate answers with citations from source documents containing ''ground truth''. A crucial factor in RAG evaluation is ''support'', or whether the information in the cited documents supports the answer. We conducted a comparative study of submissions to the TREC 2024 RAG Track, evaluating 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 good agreement between human and GPT-4o predictions. Further analysis of the disagreements shows 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. We provide a qualitative analysis of human and GPT-4o errors to help guide future evaluations.

Retrieval-augmented generation (RAG) summarizes retrieved documents into a text passage that fulfills the information need expressed by the user. Such generated responses should faithfully distill the relevant information and provide sufficient attribution back to the source documents. Nugget-based evaluation was proposed for text summarization and has been adapted to evaluate RAG output in recent shared tasks such as 2024 TREC RAG, BioGen, and NeuCLIR tracks. However, annotating such detailed and nuanced information is complex and errorful. Multiple pieces of information need to be labeled, extracted, linked, and cross-referenced. In this work, we present an annotation protocol and tool tailored to collecting information for evaluating RAG systems. Our tool has four steps: nugget creation, nugget revision, document support assessment, and finally, nugget alignment. Each step aims to minimize the annotator's cognitive load, improving the efficiency and reliability.

Information Retrieval (IR) research is deeply rooted in experimentation and evaluation, and the Text REtrieval Conference (TREC) has been playing a central role in making that possible since its inauguration in 1992. TREC's mission centers around providing the infrastructure and resources to make IR evaluations possible at scale. Over the years, a plethora of different retrieval problems were addressed, culminating in data artifacts that remained as valuable and useful tools for the IR community. Even though the data are largely available from TREC's website, there is currently no resource that facilitates a cohesive way to obtain metadata information about the run file - the IR community's de-facto standard data format for storing rankings of system-oriented IR experiments.

Large Language Models (LLMs) have enabled new ways to satisfy information needs. Although great strides have been made in applying them to settings like document ranking and short-form text generation, they still struggle to compose complete, accurate, and verifiable long-form reports. Reports with these qualities are necessary to satisfy the complex, nuanced, or multi-faceted information needs of users. In this perspective paper, we draw together opinions from industry and academia, and from a variety of related research areas, to present our vision for automatic report generation, and---critically---a flexible framework by which such reports can be evaluated. In contrast with other summarization tasks, automatic report generation starts with a detailed description of an information need, stating the necessary background, requirements, and scope of the report. Further, the generated reports should be complete, accurate, and verifiable. These qualities, which are desirable---if not required---in many analytic report-writing settings, require rethinking how to build and evaluate systems that exhibit these qualities. To foster new efforts in building these systems, we present an evaluation framework that draws on ideas found in various evaluations. To test completeness and accuracy, the framework uses nuggets of information, expressed as questions and answers, that need to be part of any high-quality generated report. Additionally, evaluation of citations that map claims made in the report to their source documents ensures verifiability.

The IARPA BETTER (Better Extraction from Text Through Enhanced Retrieval) program held three evaluations of information retrieval (IR) and information extraction (IE). For both tasks, the only training data available was in English, but systems had to perform cross-language retrieval and extraction from Arabic, Farsi, Chinese, Russian, and Korean. Pooled assessment and information extraction annotation were used to create reusable IR test collections. These datasets are freely available to researchers working in cross-language retrieval, information extraction, or the conjunction of IR and IE. This paper describes the datasets, how they were constructed, and how they might be used by researchers.

The AudioCHI 2022 workshop focusses on human engagement with spoken material in search settings, including live stream audio and collections. Spoken material comes in many forms, including for example: factual or entertaining (or both!), timely or of historical interest, local or global, single speaker or conversations. Users engage with spoken material for a variety of reasons, including entertainment, current affairs, education, and research. While there has been considerable previous work studying spoken document retrieval or more generally spoken content retrieval, AudioCHI 2022 is the first meeting to explore user engagement with audio content, including discussing: (i) how content analysis might establish verbal and non-verbal features for rich content representations, and (ii) and use cases and human factors in interaction with spoken audio content, and their interaction with more established topics relating to spoken content retrieval. The workshop brings together researchers in spoken content retrieval with expertise on human computer interaction in information access to examine opportunities and challenges for advancing technologies for search and interaction with spoken content.

Abstractive summarization of podcasts is motivated by the growing popularity of podcasts and the needs of their listeners. Podcasting is a markedly different domain from news and other media that are commonly studied in the context of automatic summarization. As such, the qualities of a good podcast summary are yet unknown. Using a collection of podcast summaries produced by different algorithms alongside human judgments of summary quality obtained from the TREC 2020 Podcasts Track, we study the correlations between various automatic evaluation metrics and human judgments, as well as the linguistic aspects of summaries that result in strong evaluations.

Rapidly growing online podcast archives contain diverse content on a wide range of topics. These archives form an important resource for entertainment and professional use, but their value can only be realized if users can rapidly and reliably locate content of interest. Search for relevant content can be based on metadata provided by content creators, but also on transcripts of the spoken content itself. Excavating relevant content from deep within these audio streams for diverse types of information needs requires varying the approach to systems prototyping. We describe a set of diverse podcast information needs and different approaches to assessing retrieved content for relevance. We use these information needs in an investigation of the utility and effectiveness of these information sources. Based on our analysis, we recommend approaches for indexing and retrieving podcast content for ad hoc search.

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.

In this paper we propose a new evaluation challenge and direction in the area of High-level Video Understanding. The challenge we are proposing is designed to test automatic video analysis and understanding, and how accurately systems can comprehend a movie in terms of actors, entities, events and their relationship to each other. A pilot High-Level Video Understanding (HLVU) dataset of open source movies were collected for human assessors to build a knowledge graph representing each of them. A set of queries will be derived from the knowledge graph to test systems on retrieving relationships among actors, as well as reasoning and retrieving non-visual concepts. The objective is to benchmark if a computer system can "understand" non-explicit but obvious relationships the same way humans do when they watch the same movies. This is long-standing problem that is being addressed in the text domain and this project moves similar research to the video domain. Work of this nature is foundational to future video analytics and video understanding technologies. This work can be of interest to streaming services and broadcasters hoping to provide more intuitive ways for their customers to interact with and consume video content.

Replicability and reproducibility of experimental results are primary concerns in all the areas of science and IR is not an exception. Besides the problem of moving the field towards more reproducible experimental practices and protocols, we also face a severe methodological issue: we do not have any means to assess when reproduced is reproduced. Moreover, we lack any reproducibility-oriented dataset, which would allow us to develop such methods.

Traditional practice recommends that information retrieval experiments be run over multiple test collections, to support, if not prove, that gains in performance are likely to generalize to other collections or tasks. However, because of the pooling assumptions, evaluation scores are not directly comparable across different test collections. We present a widely-used statistical tool, \em meta-analysis, as a framework for reporting results from IR experiments using multiple test collections. We demonstrate the meta-analytical approach through two standard experiments on stemming and pseudo-relevance feedback, and compare the results to those obtained from score standardization. Meta-analysis incorporates several recent recommendations in the literature, including score standardization, reporting effect sizes rather than score differences, and avoiding a reliance on null-hypothesis statistical testing, in a unified approach. It therefore represents an important methodological improvement over using these techniques in isolation.

This is a full-day tutorial on building and validating test collections. The intended audience is advanced students who nd themselves in need of a test collection, or actually in the process of building a test collection, to support their own research. Not everyone can talk TREC, CLEF, INEX, or NTCIR into running a track to build the collection you need. The goal of this tutorial is to lay out issues, procedures, pitfalls, and practical advice.

This article explores a method for more accurately estimating the main effect of the system in a typical test-collection-based evaluation of information retrieval systems, thus increasing the sensitivity of system comparisons. Randomly partitioning the test document collection allows for multiple tests of a given system and topic (replicates). Bootstrap ANOVA can use these replicates to extract system-topic interactions-something not possible without replicates-yielding a more precise value for the system effect and a narrower confidence interval around that value. Experiments using multiple TREC collections demonstrate that removing the topic-system interactions substantially reduces the confidence intervals around the system effect as well as increases the number of significant pairwise differences found. Further, the method is robust against small changes in the number of partitions used, against variability in the documents that constitute the partitions, and the measure of effectiveness used to quantify system effectiveness.

This paper describes a new test collection for passage retrieval from multilingual, informal text. The task being modeled is that of a monolingual English-speaking user who wishes to search discussion forum text in a foreign language. The system retrieves relevant short passages of text and presents them to the user, translated into English. The test collection contains more than 2 billion words of discussion thread text, 250 queries representing complex informational search needs, and manual relevance judgments of forum post passages, pooled from real systems. This information retrieval test collection is the first to combine multilingual search, passage retrieval, and informal online genre text.

Due to lack of mature techniques in privacy-preserving information retrieval (IR), concerns about information privacy and security have become serious obstacles that prevent valuable user data to be used in IR research such as studies on query logs, social media, and medical record retrieval. In SIGIR 2014 and SIGIR 2015, we have run the privacy-preserving IR workshops exploring and understanding the privacy and security risks in information retrieval. This year, we continue the efforts of connecting the two disciplines of IR and privacy/security by organizing this workshop. We target on three themes, differential privacy and IR dataset release, privacy in search and browsing, and privacy in social media. The workshop includes panels with researchers from both fields on these three themes, as well as invite industry speakers for real-world challenges. The goals of this workshop include (1) bringing together the two research fields, and (2) yielding fruitful collaborations.

Information retrieval (IR) and information privacy/security are two fast-growing computer science disciplines. There are many synergies and connections between these two disciplines. However, there have been very limited efforts to connect the two important disciplines. On the other hand, due to lack of mature techniques in privacy-preserving IR, concerns about information privacy and security have become serious obstacles that prevent valuable user data to be used in IR research such as studies on query logs, social media, tweets, and medical record retrieval. We propose this privacy-preserving IR workshop to connect the two disciplines of information retrieval and information privacy and security. We look forward to spurring research that aims to bring together the research fields of IR and privacy/security. Last year, the first privacy-preserving IR workshop focused on mitigating privacy threats in information retrieval by novel algorithms and tools that enable web users to better understand associated privacy risks.

While existing test collections and evaluation conference efforts may sufficiently support one's research, one can easily find oneself wanting to solve problems no one else is solving yet. But how can research in IR be done (or be published!) without solid data and experiments? Not everyone can talk TREC, CLEF, INEX, or NTCIR into running a track to build a collection.

The Twitter real-time information network is the subject of research for information retrieval tasks such as real-time search. However, so far, reproducible experimentation on Twitter data has been impeded by restrictions imposed by the Twitter terms of service. In this paper, we detail a new methodology for legally building and distributing Twitter corpora, developed through collaboration between the Text REtrieval Conference (TREC) and Twitter. In particular, we detail how the first publicly available Twitter corpus - referred to as Tweets2011 - was distributed via lists of tweet identifiers and specialist tweet crawling software. Furthermore, we analyse whether this distribution approach remains robust over time, as tweets in the corpus are removed either by users or Twitter itself. Tweets2011 was successfully used by 58 participating groups for the TREC 2011 Microblog track, while our results attest to the robustness of the crawling methodology over time.

Traditional editorial effectiveness measures, such as nDCG, remain standard for Web search evaluation. Unfortunately, these traditional measures can inappropriately reward redundant information and can fail to reflect the broad range of user needs that can underlie a Web query. To address these deficiencies, several researchers have recently proposed effectiveness measures for novelty and diversity. Many of these measures are based on simple cascade models of user behavior, which operate by considering the relationship between successive elements of a result list. The properties of these measures are still poorly understood, and it is not clear from prior research that they work as intended. In this paper we examine the properties and performance of cascade measures with the goal of validating them as tools for measuring effectiveness. We explore their commonalities and differences, placing them in a unified framework; we discuss their theoretical difficulties and limitations, and compare the measures experimentally, contrasting them against traditional measures and against other approaches to measuring novelty. Data collected by the TREC 2009 Web Track is used as the basis for our experimental comparison. Our results indicate that these measures reward systems that achieve an balance between novelty and overall precision in their result lists, as intended. Nonetheless, other measures provide insights not captured by the cascade measures, and we suggest that future evaluation efforts continue to report a variety of measures.

Recent efforts in test collection building have focused on scaling back the number of necessary relevance judgments and then scaling up the number of search topics. Since the largest source of variation in a Cranfield-style experiment comes from the topics, this is a reasonable approach. However, as topic set sizes grow, and researchers look to crowdsourcing and Amazon's Mechanical Turk to collect relevance judgments, we are faced with issues of quality control. This paper examines the robustness of the TREC Million Query track methods when some assessors make significant and systematic errors. We find that while averages are robust, assessor errors can have a large effect on system rankings.

In opinion-finding, the retrieval system is tasked with retrieving not just relevant documents, but those that also express an opinion towards the query target entity. This task has been studied in the context of the blogosphere by groups participating in the 2006-2008 TREC Blog tracks. Spam blogs (splogs) are thought to be a problem on the blogosphere. In this paper, we investigate the extent to which spam has affected the participating groups' retrieval systems over the three years of the TREC Blog track opinion-finding task. Our results show that spam can be an issue, with most systems retrieving some spam for every topic. However, removing spam from the rankings does not markedly change the relative performance of opinion-finding approaches.

We investigate to what extent people making relevance judgements for a reusable IR test collection are exchangeable. We consider three classes of judge: "gold standard" judges, who are topic originators and are experts in a particular information seeking task; "silver standard" judges, who are task experts but did not create topics; and "bronze standard" judges, who are those who did not define topics and are not experts in the task.

In opinion-finding, the retrieval system is tasked with retrieving not just relevant documents, but which also express an opinion towards the query target entity. Most opinion-finding systems are based on a two-stage approach, where initially the system aims to retrieve relevant documents, which are then re-ranked according to the extent to which they are detected to be of an opinionated nature. In this work, we investigate how the underlying 'baseline' retrieval system performance affects the overall opinion-finding performance. We apply two effective opinion-finding techniques to all the baseline runs submitted to the TREC 2007 Blog track, and draw new insights and conclusions.

Information retrieval evaluation based on the pooling method is inherently biased against systems that did not contribute to the pool of judged documents. This may distort the results obtained about the relative quality of the systems evaluated and thus lead to incorrect conclusions about the performance of a particular ranking technique.

This poster describes a potential problem with a relatively well used measure in Information Retrieval research: Kendall's Tau rank correlation coefficient. The coefficient is best known for its use in determining the similarity of test collections when ranking sets of retrieval runs. Threshold values for the coefficient have been defined and used in a number of published studies in information retrieval. However, this poster presents results showing that basing decisions on such thresholds is not as reliableas has been assumed.

Test collections are most useful when they are reusable, that is, when they can be reliably used to rank systems that did not contribute to the pools. Pooled relevance judgments for very large collections may not be reusable for two easons: they will be very sparse and not sufficiently complete, and they may be biased in the sense that theywill unfairly rank some class of systems. The TREC 2006 terabyte track judged both a pool and a deep random sample in order to measure the effects of sparseness and bias.

Modern retrieval test collections are built through a process called pooling in which only a sample of the entire document set is judged for each topic. The idea behind pooling is to find enough relevant documents such that when unjudged documents are assumed to be nonrelevant the resulting judgment set is sufficiently complete and unbiased. As document sets grow larger, a constant-size pool represents an increasingly small percentage of the document set, and at some point the assumption of approximately complete judgments must become invalid.This paper demonstrates that the AQUAINT 2005 test collection exhibits bias caused by pools that were too shallow for the document set size despite having many diverse runs contribute to the pools. The existing judgment set favors relevant documents that contain topic title words even though relevant documents containing few topic title words are known to exist in the document set. The paper concludes with suggested modifications to traditional pooling and evaluation methodology that may allow very large reusable test collections to be built.

Existing methods for measuring the quality of search algorithms use a static collection of documents. A set of queries and a mapping from the queries to the relevant documents allow the experimenter to see how well different search engines or engine configurations retrieve the correct answers. This methodology assumes that the document set and thus the set of relevant documents are unchanging. In this paper, we abandon the static collection requirement. We begin with a recent TREC collection created from a web crawl and analyze how the documents in that collection have changed over time. We determine how decay of the document collection affects TREC systems, and present the results of an experiment using the decayed collection to measure a live web search system. We employ novel measures of search effectiveness that are robust despite incomplete relevance information. Lastly, we propose a methodology of "collection maintenance" which supports measuring search performance both for a single system and between systems run at different points in time.

Many common web searches by their nature have a very small number of relevant documents. Homepage and "namedpage" searching are known-item searches where there is only a single relevant document. Topic distillation is a special kind of topical relevance search where the user wishes to find a few key web sites rather than every relevant web page. Because these types of searches are so common, web search evaluations have come to focus on tasks where there are very few relevant documents. Evaluations with few relevant documents pose special challenges for current metrics. In particular, the TREC 2003 topic distillation evaluation is unable to distinguish most submitted runs from each other.

Test collections for the filtering track in TREC have typically used either past sets of relevance judgments, or categorized collections such as Reuters Corpus Volume 1 or OHSUMED, because filtering systems need relevance judgments during the experiment for training and adaptation. For TREC 2002, we constructed an entirely new set of search topics for the Reuters Corpus for measuring filtering systems. Our method for building the topics involved multiple iterations of feedback from assessors, and fusion of results from multiple search systems using different search algorithms. We also developed a second set of inexpensive topics based on categories in the document collection. We found that the initial judgments made for the experiment were sufficient; subsequent pooled judging changed system rankings very little. We also found that systems performed very differently on the category topics than on the assessor-built topics.

Users of the World-Wide Web are not only confronted by an immense overabundance of information, but also by a plethora of tools for searching for the web pages that suit their information needs. Web search engines differ widely in interface, features, coverage of the web, ranking methods, delivery of advertising, and more. In this paper, we present a method for comparing search engines automatically based on how they rank known item search results. Because the engines perform their search on overlapping (but different) subsets of the web collected at different points in time, evaluation of search engines poses significant challenges to the traditional information retrieval methodology. Our method uses known item searching; comparing the relative ranks of the items in the search engines' rankings. Our approach automatically constructs known item queries using query log analysis and automatically constructs the result via analysis of editor comments from the ODP (Open Directory Project). Additionally, we present our comparison on five (Lycos, Netscape, Fast, Google, HotBot) well-known search services and find that some services perform known item searches better than others, but the majority are statistically equivalent.

We measure the WT10g test collection, used in the TREC-9 and TREC 2001 Web Tracks, with common measures used in the web topology community, in order to see if WT10g "looks like" the web. This is not an idle question; characteristics of the web, such as power law relationships, diameter, and connected components have all been observed within the scope of general web crawls, constructed by blindly following links. In contrast, WT10g was carved out from a larger crawl specifically to be a web search test collection within the reach of university researchers. Does such a collection retain the properties of the larger web? In the case of WT10g, yes.

The most prevalent experimental methodology for comparing the effectiveness of information retrieval systems requires a test collection, composed of a set of documents, a set of query topics, and a set of relevance judgments indicating which documents are relevant to which topics. It is well known that relevance judgments are not infallible, but recent retrospective investigation into results from the Text REtrieval Conference (TREC) has shown that differences in human judgments of relevance do not affect the relative measured performance of retrieval systems. Based on this result, we propose and describe the initial results of a new evaluation methodology which replaces human relevance judgments with a randomly selected mapping of documents to topics which we refer to aspseudo-relevance judgments.Rankings of systems with our methodology correlate positively with official TREC rankings, although the performance of the top systems is not predicted well. The correlations are stable over a variety of pool depths and sampling techniques. With improvements, such a methodology could be useful in evaluating systems such as World-Wide Web search engines, where the set of documents changes too often to make traditional collection construction techniques practical.

Collaborative filtering is a technique for recommending documents to users based on how similar their tastes are to other users. If two users tend to agree on what they like, the system will recommend the same documents to them. The generalized vector space model of information retrieval represents a document by a vector of its similarities to all other documents. The process of collaborative filtering is nearly identical to the process of retrieval using GVSM in a matrix of user ratings. Using this observation, a model for filtering collaboratively using document content is possible.