Search

Cluster retrieval assumes that the probability of relevance of a document should depend on the relevance of other similar documents to the same query. The goal is to find the best group of documents. Many studies have examined the effectiveness of this approach, by employing different retrieval methods or clustering algorithms, but few have investigated text representations. This paper revisits the problem of retrieving the best group of documents, from the language-modeling perspective. We analyze the advantages and disadvantages of a range of representation techniques, derive features that characterize the good document groups, and experiment with a new probabilistic representation as a first step toward incorporating these features. Empirical evaluation demonstrates that the relationship between documents can be leveraged in retrieval when a good representation technique is available, and that retrieving the best group of documents can be more effective than retrieving individual documents. [ABSTRACT FROM AUTHOR]

Many traditional information retrieval models, such as BM25 and language modeling, give good retrieval effectiveness, but can be difficult to implement efficiently. Recently, document-centric impact models were developed in order to overcome some of these efficiency issues. However, such models have a number of problems, including poor effectiveness, and heuristic term weighting schemes. In this work, we present a statistical view of document-centric impact models. We describe how such models can be treated statistically and propose a supervised parameter estimation technique. We analyze various theoretical and practical aspects of the model and show that weights estimated using our new estimation technique are significantly better than the integer-based weights used in previous studies. [ABSTRACT FROM AUTHOR]

Previous research has shown that using term associations could improve the effectiveness of information retrieval (IR) systems. However, most of the existing approaches focus on query reformulation. Document reformulation has just begun to be studied recently. In this paper, we study how to utilize term association measures to do document modeling, and what types of measures are effective in document language models. We propose a probabilistic term association measure, compare it to some traditional methods, such as the similarity co-efficient and window-based methods, in the language modeling (LM) framework, and show that significant improvements over query likelihood (QL) retrieval can be obtained. We also compare the method with state-of-the-art document modeling techniques based on latent mixture models. [ABSTRACT FROM AUTHOR]

We investigate the effectiveness of query structuring in the Japanese language by composing or decomposing compound words and phrases. Our method is based on a theoretical framework using Markov random fields. Our two-stage term dependence model captures both the global dependencies between query components explicitly delimited by separators in a query, and the local dependencies between constituents within a compound word when the compound word appears in a query component. We show that our model works well, particularly when using query structuring with compound words, through experiments using a 100-gigabyte web document collection mostly written in Japanese. [ABSTRACT FROM AUTHOR]

The chapter presented a framework for understanding and studying how behavior can be used as implicit relevance feedback. This included a classification and discussion of behaviors that have been used as implicit relevance feedback, a general discussion and characterization of implicit feedback research, and a presentation of example studies to illustrate how such studies have been conducted and how feedback has typically been measured and used. Finally, this chapter presented a discussion of key issues and problems associated with implicit feedback research and identified challenges for future research. The use of behavior as implicit relevance feedback is an exciting and promising approach to personalizing IR interactions. Although more effort needs to be made to fully understand how behaviors can be used as implicit relevance feedback, current research efforts offer a promising start. [ABSTRACT FROM AUTHOR]

At present no overall integrative framework exists for CIR, but an integrated approach—i.e., that distinguishes separate concepts and processes in information seeking and information search then attempts to create an integrated conceptualization of the user who is both searching and seeking information within the larger context of HIB—has the potential to yield a more holistic theoretical and cognitive understanding that will assist IR and Web system designers. This book provides an overview of new directions in CIR research. The field of CIR is broad, international, interdisciplinary and dynamic with tremendous potential to impact the everyday lives of people in both developed and developing countries as they increasingly need to interact with IR systems. This book is not an exhaustive or historical discussion of all possible areas of important and new directions in CIR research. Information retrieval, in all its technical, cognitive and other respects continues to be an intractable research problem and research area. Our goal in producing this book was to stimulate the thinking of authors and readers alike. [ABSTRACT FROM AUTHOR]

Developing and supporting human search capabilities is at least equally important as developing the capabilities of search engines. Existing academic research suggests that particularly inexperienced users searching the Web utilize only a modest subset of the capabilities that the tools offer and are weak at understanding their real information needs, articulating them in a way that allows for effective searches, and interpreting search results in the context of those needs. This clearly suggests that training users to become better searchers is a worthwhile effort, and that understanding what makes certain search interventions successful and others not is vitally important for enabling users to make effective use of their time. Our review of existing search training literature revealed relatively broadly scattered efforts that, to a large extent, were not programmatic or focused on developing well-defined theoretical models. This makes it difficult to integrate findings from different studies into cohesive bodies of work that would support the development and evaluation of new training interventions. In this chapter, we propose three approaches to using existing areas of research as a theoretical foundation for future work. Namely: a) understanding the nature of search processes is very important if we want to understand search training, because different stages of the search process require different training approaches; b) a thorough analysis of search errors forms a solid foundation for training approaches that helps users avoid the common errors; and c) training research from related disciplines, such as information systems, can be used to introduce new theoretical perspectives and suggest models that are also potentially helpful in search training. Search is pervasive in modern life and performing effective online searches is a fundamentally important literacy skill, not only for 21st century knowledge professionals but for every individual who relies on material available from online sources. Understanding how we, as a research community, can help people become effective searchers is, therefore, of critical importance. [ABSTRACT FROM AUTHOR]

The colossal landscape of scholarly knowledge, growing at exponential rates, now requires representational maps utilizing advanced techniques to provide insight into the structure and dynamics of scholarly domains. Today we need intellectual cartographers to assist students and scholars in navigating, understanding and internalizing the structure and dynamics of scholarly bodies of knowledge. There is compelling evidence of the utility of KDVs stemming from the fields of educational psychology, cognitive science, cartography, and information science. Well designed KDVs have the ability to facilitate understanding, recall, and to convey to the user the schematic, geo-spatial, f temporal, semantic, or social organization of the underlying domain. Though educational knowledge domain visualizations are still in their infancy, we believe that they have a promising future in assisting with access to and the navigation, understanding, management, and communication of large-scale information spaces. Furthermore, when used as an interface for information retrieval, knowledge domain visualizations have the potential to convey the structural organization of the domain to the user. In turn, this structural knowledge of the domain provides the cognitive scaffolding with which the user may associate additional details about the domain. [ABSTRACT FROM AUTHOR]

The principle of polyrepresentation is a coherent and comprehensive cognitive framework that can be applied simultaneously to the cognitive space of the user and the information space of IR systems. The principle has the potential to guide the design of interactive IR systems that take full advantage of the available document representations and user’s context to improve retrieval performance. However, before this can be achieved a number of issues need further investigation. Among these are simulation studies that test which methods would be appropriate for matching different representations of the user’s cognitive space with document representations. Such simulations could apply simulated work task situations; or they could explore exhaustively the possibilities of a number of controlled variables and thus simulate all achievable combinations. Investigations involving test persons and experimental laboratory tests (simulations) must take into account the dependency of domains, media and representation styles. Studies of the principles for how a Request Model Builder should function would be fruitful, and how to match representations to generate strong cognitive overlaps, especially in best match settings. The latter issue is illustrated by the polyrepresentation continuum. It points to the investigation of flexible and powerful hybrid matching of representations as a challenge and opportunity for future research along these lines. [ABSTRACT FROM AUTHOR]