Your search

Learning to rank for Information Retrieval (IR) is a task to automatically construct a ranking model using training data, such that the model can sort new objects according to their degrees of relevance, preference, or importance. Many IR problems are by nature ranking problems, and many IR technologies can be potentially enhanced by using learning-to-rank techniques. The objective of this tutorial is to give an introduction to this research direction. Specifically, the existing learning-to-rank algorithms are reviewed and categorized into three approaches: the pointwise, pairwise, and listwise approaches. The advantages and disadvantages with each approach are analyzed, and the relationships between the loss functions used in these approaches and IR evaluation measures are discussed. Then the empirical evaluations on typical learning-to-rank methods are shown, with the LETOR collection as a benchmark dataset, which seems to suggest that the listwise approach be the most effective one among all the approaches. After that, a statistical ranking theory is introduced, which can describe different learning-to-rank algorithms, and be used to analyze their query-level generalization abilities. At the end of the tutorial, we provide a summary and discuss potential future work on learning to rank.

This paper examines the reliability of implicit feedback generated from clickthrough data in WWW search. Analyzing the users' decision process using eyetracking and comparing implicit feedback against manual relevance judgments, we conclude that clicks are informative but biased. While this makes the interpretation of clicks as absolute relevance judgments difficult, we show that relative preferences derived from clicks are reasonably accurate on average.

Previous work on understanding user web search behavior has focused on how people search and what they are searching for, but not why they are searching. In this paper, we describe a framework for understanding the underlying goals of user searches, and our experience in using the framework to manually classify queries from a web search engine. Our analysis suggests that so-called navigational" searches are less prevalent than generally believed while a previously unexplored "resource-seeking" goal may account for a large fraction of web searches. We also illustrate how this knowledge of user search goals might be used to improve future web search engines.

This paper presents an approach to automatically optimizing the retrieval quality of search engines using clickthrough data. Intuitively, a good information retrieval system should present relevant documents high in the ranking, with less relevant documents following below. While previous approaches to learning retrieval functions from examples exist, they typically require training data generated from relevance judgments by experts. This makes them difficult and expensive to apply. The goal of this paper is to develop a method that utilizes clickthrough data for training, namely the query-log of the search engine in connection with the log of links the users clicked on in the presented ranking. Such clickthrough data is available in abundance and can be recorded at very low cost. Taking a Support Vector Machine (SVM) approach, this paper presents a method for learning retrieval functions. From a theoretical perspective, this method is shown to be well-founded in a risk minimization framework. Furthermore, it is shown to be feasible even for large sets of queries and features. The theoretical results are verified in a controlled experiment. It shows that the method can effectively adapt the retrieval function of a meta-search engine to a particular group of users, outperforming Google in terms of retrieval quality after only a couple of hundred training examples.

In studying actual Web searching by the public at large, we analyzed over one million Web queries by users of the Excite search engine. We found that most people use few search terms, few modified queries, view few Web pages, and rarely use advanced search features. A small number of search terms are used with high frequency, and a great many terms are unique; the language of Web queries is distinctive. Queries about recreation and entertainment rank highest. Findings are compared to data from two other large studies of Web queries. This study provides an insight into the public practices and choices in Web searching.

We analyzed transaction logs containing 51,473 queries posed by 18,113 users of Excite, a major Internet search service. We provide data on: (i) sessions — changes in queries during a session, number of pages viewed, and use of relevance feedback; (ii) queries — the number of search terms, and the use of logic and modifiers; and (iii) terms — their rank/frequency distribution and the most highly used search terms. We then shift the focus of analysis from the query to the user to gain insight to the characteristics of the Web user. With these characteristics as a basis, we then conducted a failure analysis, identifying trends among user mistakes. We conclude with a summary of findings and a discussion of the implications of these findings.

Of growing interest in the area of improving the search experience is the collection of implicit user behavior measures (implicit measures) as indications of user interest and user satisfaction. Rather than having to submit explicit user feedback, which can be costly in time and resources and alter the pattern of use within the search experience, some research has explored the collection of implicit measures as an efficient and useful alternative to collecting explicit measure of interest from users.This research article describes a recent study with two main objectives. The first was to test whether there is an association between explicit ratings of user satisfaction and implicit measures of user interest. The second was to understand what implicit measures were most strongly associated with user satisfaction. The domain of interest was Web search. We developed an instrumented browser to collect a variety of measures of user activity and also to ask for explicit judgments of the relevance of individual pages visited and entire search sessions. The data was collected in a workplace setting to improve the generalizability of the results.Results were analyzed using traditional methods (e.g., Bayesian modeling and decision trees) as well as a new usage behavior pattern analysis (“gene analysis”). We found that there was an association between implicit measures of user activity and the user's explicit satisfaction ratings. The best models for individual pages combined clickthrough, time spent on the search result page, and how a user exited a result or ended a search session (exit type/end action). Behavioral patterns (through the gene analysis) can also be used to predict user satisfaction for search sessions.

The use of data stored in transaction logs of Web search engines, Intranets, and Web sites can provide valuable insight into understanding the information-searching process of online searchers. This understanding can enlighten information system design, interface development, and devising the information architecture for content collections. This article presents a review and foundation for conducting Web search transaction log analysis. A methodology is outlined consisting of three stages, which are collection, preparation, and analysis. The three stages of the methodology are presented in detail with discussions of goals, metrics, and processes at each stage. Critical terms in transaction log analysis for Web searching are defined. The strengths and limitations of transaction log analysis as a research method are presented. An application to log client-side interactions that supplements transaction logs is reported on, and the application is made available for use by the research community. Suggestions are provided on ways to leverage the strengths of, while addressing the limitations of, transaction log analysis for Web-searching research. Finally, a complete flat text transaction log from a commercial search engine is available as supplementary material with this manuscript.

Evaluation measures act as objective functions to be optimized by information retrieval systems. Such objective functions must accurately reflect user requirements, particularly when tuning IR systems and learning ranking functions. Ambiguity in queries and redundancy in retrieved documents are poorly reflected by current evaluation measures. In this paper, we present a framework for evaluation that systematically rewards novelty and diversity. We develop this framework into a specific evaluation measure, based on cumulative gain. We demonstrate the feasibility of our approach using a test collection based on the TREC question answering track.

In this paper, we define and present a comprehensive classification of user intent for Web searching. The classification consists of three hierarchical levels of informational, navigational, and transactional intent. After deriving attributes of each, we then developed a software application that automatically classified queries using a Web search engine log of over a million and a half queries submitted by several hundred thousand users. Our findings show that more than 80% of Web queries are informational in nature, with about 10% each being navigational and transactional. In order to validate the accuracy of our algorithm, we manually coded 400 queries and compared the results from this manual classification to the results determined by the automated method. This comparison showed that the automatic classification has an accuracy of 74%. Of the remaining 25% of the queries, the user intent is vague or multi-faceted, pointing to the need for probabilistic classification. We discuss how search engines can use knowledge of user intent to provide more targeted and relevant results in Web searching.

Understanding user intent is key to designing an effective ranking system in a search engine. In the absence of any explicit knowledge of user intent, search engines want to diversify results to improve user satisfaction. In such a setting, the probability ranking principle-based approach of presenting the most relevant results on top can be sub-optimal, and hence the search engine would like to trade-off relevance for diversity in the results. In analogy to prior work on ranking and clustering systems, we use the axiomatic approach to characterize and design diversification systems. We develop a set of natural axioms that a diversification system is expected to satisfy, and show that no diversification function can satisfy all the axioms simultaneously. We illustrate the use of the axiomatic framework by providing three example diversification objectives that satisfy different subsets of the axioms. We also uncover a rich link to the facility dispersion problem that results in algorithms for a number of diversification objectives. Finally, we propose an evaluation methodology to characterize the objectives and the underlying axioms. We conduct a large scale evaluation of our objectives based on two data sets: a data set derived from the Wikipedia disambiguation pages and a product database.

This paper provides overview and instruction regarding the evaluation of interactive information retrieval systems with users. The primary goal of this article is to catalog and compile material related to this topic into a single source. This article (1) provides historical background on the development of user-centered approaches to the evaluation of interactive information retrieval systems; (2) describes the major components of interactive information retrieval system evaluation; (3) describes different experimental designs and sampling strategies; (4) presents core instruments and data collection techniques and measures; (5) explains basic data analysis techniques; and (4) reviews and discusses previous studies. This article also discusses validity and reliability issues with respect to both measures and methods, presents background information on research ethics and discusses some ethical issues which are specific to studies of interactive information retrieval (IIR). Finally, this article concludes with a discussion of outstanding challenges and future research directions.

This study examined how searchers interacted with a web-based, faceted library catalog when conducting exploratory searches. It applied eye tracking, stimulated recall interviews, and direct observation to investigate important aspects of gaze behavior in a faceted search interface: what components of the interface searchers looked at, for how long, and in what order. It yielded empirical data that will be useful for both practitioners (e.g., for improving search interface designs), and researchers (e.g., to inform models of search behavior). Results of the study show that participants spent about 50 seconds per task looking at (fixating on) the results, about 25 seconds looking at the facets, and only about 6 seconds looking at the query itself. These findings suggest that facets played an important role in the exploratory search process.

The goal of the Redundancy, Diversity, and Interdependent Document Relevance workshop was to explore how ranking, performance assessment and learning to rank can move beyond the assumption that the relevance of a document is independent of other documents. In particular, the workshop focussed on three themes: the effect of redundancy on information retrieval utility (for example, minimizing the wasted effort of users who must skip redundant information), the role of diversity (for example, for mitigating the risk of misinterpreting ambiguous queries), and algorithms for set-level optimization (where the quality of a set of retrieved documents is not simply the sum of its parts). This workshop built directly upon the Beyond Binary Relevance: Preferences, Diversity and Set-Level Judgments workshop at SIGIR 2008 [3], shifting focus to address the questions left open by the discussions and results from that workshop. As such, it was the first workshop to explicitly focus on the related research challenges of redundancy, diversity, and interdependent relevance – all of which require novel performance measures, learning methods, and evaluation techniques. The workshop program committee consisted of 15 researchers from academia and industry, with experience in IR evaluation, machine learning, and IR algorithmic design. Over 40 people attended the workshop. This report aims to summarize the workshop, and also to systematize common themes and key concepts so as to encourage research in the three workshop themes. It contains our attempt to summarize and organize the topics that came up in presentations as well as in discussions, pulling out common elements. Many audience members contributed, yet due to the free-flowing discussion, attributing all the observations to particular audience members is unfortunately impossible. Not all audience members would necessarily agree with the views presented, but we do attempt to present a consensus view as far as possible.

The Probabilistic Relevance Framework (PRF) is a formal framework for document retrieval, grounded in work done in the 1970–1980s, which led to the development of one of the most successful text-retrieval algorithms, BM25. In recent years, research in the PRF has yielded new retrieval models capable of taking into account document meta-data (especially structure and link-graph information). Again, this has led to one of the most successful Web-search and corporate-search algorithms, BM25F. This work presents the PRF from a conceptual point of view, describing the probabilistic modelling assumptions behind the framework and the different ranking algorithms that result from its application: the binary independence model, relevance feedback models, BM25 and BM25F. It also discusses the relation between the PRF and other statistical models for IR, and covers some related topics, such as the use of non-textual features, and parameter optimisation for models with free parameters.

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.

In this paper we describe a general framework for evaluation and optimization of methods for diversifying query results. In these methods, an initial ranking candidate set produced by a query is used to construct a result set, where elements are ranked with respect to relevance and diversity features, i.e., the retrieved elements should be as relevant as possible to the query, and, at the same time, the result set should be as diverse as possible. While addressing relevance is relatively simple and has been heavily studied, diversity is a harder problem to solve. One major contribution of this paper is that, using the above framework, we adapt, implement and evaluate several existing methods for diversifying query results. We also propose two new approaches, namely the Greedy with Marginal Contribution (GMC) and the Greedy Randomized with Neighborhood Expansion (GNE) methods. Another major contribution of this paper is that we present the first thorough experimental evaluation of the various diversification techniques implemented in a common framework. We examine the methods' performance with respect to precision, running time and quality of the result. Our experimental results show that while the proposed methods have higher running times, they achieve precision very close to the optimal, while also providing the best result quality. While GMC is deterministic, the randomized approach (GNE) can achieve better result quality if the user is willing to tradeoff running time.

As information becomes more ubiquitous and the demands that searchers have on search systems grow, there is a need to support search behaviors beyond simple lookup. Information seeking is the process or activity of attempting to obtain information in both human and technological contexts. Exploratory search describes an information-seeking problem context that is open-ended, persistent, and multifaceted, and information-seeking processes that are opportunistic, iterative, and multitactical. Exploratory searchers aim to solve complex problems and develop enhanced mental capacities. Exploratory search systems support this through symbiotic human-machine relationships that provide guidance in exploring unfamiliar information landscapes. Exploratory search has gained prominence in recent years. There is an increased interest from the information retrieval, information science, and human-computer interaction communities in moving beyond the traditional turn-taking interaction model supported by major Web search engines, and toward support for human intelligence amplification and information use. In this lecture, we introduce exploratory search, relate it to relevant extant research, outline the features of exploratory search systems, discuss the evaluation of these systems, and suggest some future directions for supporting exploratory search. Exploratory search is a new frontier in the search domain and is becoming increasingly important in shaping our future world.

In 1975 Tefko Saracevic declared “the subject knowledge view” to be the most fundamental perspective of relevance. This paper examines the assumptions in different views of relevance, including “the system's view” and “the user's view” and offers a reinterpretation of these views. The paper finds that what was regarded as the most fundamental view by Saracevic in 1975 has not since been considered (with very few exceptions). Other views, which are based on less fruitful assumptions, have dominated the discourse on relevance in information retrieval and information science. Many authors have reexamined the concept of relevance in information science, but have neglected the subject knowledge view, hence basic theoretical assumptions seem not to have been properly addressed. It is as urgent now as it was in 1975 seriously to consider “the subject knowledge view” of relevance (which may also be termed “the epistemological view”). The concept of relevance, like other basic concepts, is influenced by overall approaches to information science, such as the cognitive view and the domain-analytic view. There is today a trend toward a social paradigm for information science. This paper offers an understanding of relevance from such a social point of view.