Croft, W. Bruce, & Raj Das. 1990. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the Thirteenth International ACM/SIGIR Conference, 349--365.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... : Most methods for deriving classification functions from natural language queries use statistics of word occurrences to set the coefficients of a linear discriminant function [5, 20] The best results are obtained when supervised machine learning, in the guise of relevance feedback, is used [21, 6]. Text categorization systems can also be viewed as computing a function defined over documents, in this case a k ary function, where k is the number of categories into which documents can be sorted. Rather than deriving this function from a natural language query, it is typically constructed ....

....often turned out to be unreliable, so the larger collections are preferred. Evaluation is still a research issue in IR. The exhaustive relevance judgments assumed for traditional test collections are not possible with larger collections, nor when evaluating highly interactive retrieval systems [6]. For more on evaluation in IR, the reader is referred to Sparck Jones excellent collection on the subject [25] Evaluation of text categorization systems also needs more attention. One approach is to treat each category as a query and compute average recall and precision across categories [12] ....

[Article contains additional citation context not shown here]

W. Bruce Croft and Raj Das. Experiments with query acquisition and use in document retrieval systems. In Thirteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 1990. To appear.

....user inputs about concepts related to those mentioned in CHAPTER 2. RELATED WORK 10 an initial query, together with their relative importance, can significantly improve retrieval effectiveness [SB98] Relevance feedback can be improved if users select features from the texts of relevant documents [CD90] instead of limiting them to selecting concepts from lists of terms selected automatically from relevant documents by the system (the system extracts these concepts by applying natural language processing techniques to the descriptions of interests that the user provides) The techniques for ....

W.B. Croft and Das. Experiments with query acquisition and use in document retrieval system. In ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368. ACM, 1990.

....lost in conceptual space. This problem of feeling lost in conceptual space, was found to be similar to the problem of lost in hyperspace as it can be encountered in the world of information retrieval. In the context of information retrieval, this problem is approached by using query by navigation [29, 46, 20, 1, 8]. The query by navigation interaction mechanism between a searcher and the system is well known from the Information Retrieval field, and has proven to be useful. This has triggered us in applying this principle to query formulation on information systems [16, 38, 37] The effectiveness of query ....

W.B. Croft and R. Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In J. Vidick, editor, Proceedings of the 13th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 349--368. ACM Press (1990).

....disclosure. Quite often, the quest from users to fulfill their information need can be aptly described by ( Bru93] 2 I don t know what I m looking for, but I ll know when I find it. In document retrieval systems this problem is attacked by using query by navigation ( GGP89] Luc90] CD90] ACG91] BW92] and relevance feedback mechanisms ( Rij89] The query by navigation interaction mechanism between a searcher and the system is well known from the Information Retrieval field, and has proven to be useful. The above discussed parallel between information disclosure in the ....

W.B. Croft and R. Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In J. Vidick, editor, Proceedings of the 13th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368. ACM Press, 1990.

....Their research showed that similar terms identified by symmetric co occurrence function tended to occur very frequently in the database that is being searched and thus did little or nothing to improve the discriminatory power of the original query. ffl Relevance feedback principle: Croft and Das [16] reported significant improvements in effectiveness of expanded queries when users are prompted for additional terms that can be used in the search. Automatic term replacement or switching is often misleading and impractical, considering the unique context and backgrounds that different searchers ....

W. B. Croft and R. Das. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the 13th Conference on Research and Development in Information Retrieval, pages 349--365, Brussels, Belgium, 5-7 September 1990.

....problems [14, 15] Thirdly and most important, many end users have little skill or experience in formulating initial search requests, or in modifying their requests after observing failures. Even when relevance feedback is available, it still needs to be leveraged from a sensible starting point [6]. Thus, while established research results show that natural language indexing and searching is effective to a degree, it is natural to ask whether it is possible to improve on the very simple strategies described earlier without increasing the load on the user. More discriminating methods may be ....

Croft, W. B. and Das, R. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the Thirteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (Sep. 5--7, Brussels). ACM/SIGIR, New York, 1990, pp. 349--365.

....hope of achieving improved recall and precision. An RF module uses a selection metric to extract a set of terms from the relevant texts. The top n terms are then weighted according to another metric. For our experiments, we apply the selection and weighting metrics used in a similar application [Croft and Das, 1990] . A query consists of a weighted sum of terms. Ordinarily INQUERY would not engage in relevance feedback until a retrieval, based on user input, had been made and a set of documents retrieved, examined, and tagged by the user. However, since the CBR analysis already provides the system with a set ....

....as might be expected. In fact, in most instances our system achieved results as good as or better than with queries with fewer terms. Thus, not only is there limited cost associated with using this many terms, there is no detrimental effect. Our results stand in contrast to those of Croft and Das, [Croft and Das, 1990] , who claimed that relevance feedback may not be beneficial when using only a small set of relevant documents. We found this not to be the case. Their doubts are due to the potential lack of concept coverage by a small set of documents. However, their documents were relatively short; they used ....

W. Bruce Croft and Raj Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In 13th International Conference on Research and Development in InformationRetrieval, pp 349-- 365, 1990.

....and problems [13] Thirdly and most important, many end users have little skill or experience in formulating initial search requests, or in modifying their requests after observing failures. Even when relevance feedback is available, it still needs to be leveraged from a sensible starting point [5]. Thus, while established research results show that natural language indexing and searching is effective to a degree, it is natural to ask first, whether it is possible to improve on the very simple strategies described earlier without increasing the load on the user, and second, whether it is ....

Croft, W. B. and Das, R. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the Thirteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (Sep. 5--7, Brussels). ACM/SIGIR, New York, 1990, pp. 349--365.

....in fact, in most instances achieved results as good as or better than queries with fewer terms. Thus, not only is there limited cost associatedwith using this many terms, there is no detrimental effect. 4 In re Rasmussen, 888 F. 2d 703 These results stand in contrast to those of Croft and Das, [5], who found that relevance feedback may not be beneficial when using a small set of relevant documents. We found this not to be the case. Their belief is due to the potential lack of conceptcoverage by a small set of documents. However, their documentswere relatively short; they used abstracts ....

W. Bruce Croft and Raj Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In 13th International Conference on Research and Development in Information Retrieval, pages 349--365, 1990.

....disclosure. Quite often, the quest from users to fulfill their information need can be aptly described by ( Bru93] I don t know what I m looking for, but I ll know when I find it. In document retrieval systems this problem is attacked by using query by navigation ( GGP89] Luc90] CD90] ACG91] BW92] and relevance feedback mechanisms ( Rij89] The query by navigation interaction mechanism between a searcher and the system is well known from the Information Retrieval field, and has proven to be useful. The above discussed parallel between information disclosure in the ....

W.B. Croft and R. Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In J. Vidick, editor, Proceedings of the 13th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368. ACM Press, 1990.

....1 Introduction One approach to improving the effectiveness of an information retrieval (IR) system is to use sophisticated methods of gathering and representing information from a user. Techniques include automatic or interactive introduction of synonyms [Har88] forms based interfaces [CD90] automatic recognition of phrases [CTL91] and relevance feedback [SB90] All of these techniques have shown promise on standard test collections, but it was not clear how they would scale up to much larger and more heterogeneous document collections. A large and heterogeneous document ....

....The research results obtained are suggestive of what might work well in an interactive interface with a human user. Previous research with other collections suggested that combining different representations of an information need can yield an improvement in both recall and precision [KMT 82; CD90; Tur91] However, it has been difficult to do systematic research on this subject because different representations of information needs have not been available generally. The TIPSTER project also distinguishes among two different types of queries: AdHoc and Routing [Har92a] AdHoc queries are ....

[Article contains additional citation context not shown here]

W. B. Croft and R. Das. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368, 1990.

....assigned, while the higher quality phrase oper system occurs in 59 documents but only under 78 categories. Of course, as with clustering, a large text corpus is needed to obtain this distributional information. Another potential source of high quality phrases is the user of the IR system [6]. While the user cannot control which phrases take part in clusters, he or she can control which phrases are extracted from the query, and thus used to match clusters. If we restrict the phrases used in the CACM queries to ones identified by a human as meaningful 3 , the performance of phrases ....

....probabilistic retrieval models are inadequate for use with phrases and clusters, particularly in handling the known dependencies between terms and phrases and terms and clusters. Network models [38] and probabilistic models incorporating explicit dependencies are two promising alternatives [6]. 3 The set of phrases used was generated by a graduate student who was not involved in the experiments on syntactic phrase formation. 7 Conclusions Term clustering is a natural approach to remedying the poor statistical properties of syntactic phrases. Our preliminary experiments offer some ....

W. Bruce Croft and Raj Das. Experiments with query acquisition and use in document retrieval systems. In Thirteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 1990.

....it would be straightforward to add weights to terms in Boolean queries, a open research issue is whether significant improvements can be obtained by weighting Boolean query terms. While the Boolean queries available with the standard test collections do not include term importance information, [CD90] suggests that useful information about term importance can be readily obtained from users; this information could be directly incorporated as weights for Boolean expression evaluation. It is also possible that some form of idf weighting could be used to improve performance of Boolean queries or ....

Croft, W. Bruce and Das, Raj. Experiments with query acquisition and use in document retrieval systems. In Vidick, Jean-Luc, editor, Proceedings of the 13 th International Conference on Research and Development in Information Retrieval, pages 349--368. ACM, September 1990.

....been developed which rank the retrieved documents by a score which is based on the probability of relevance of the document to the query. The best known of such models are the vector space and probabilistic retrieval models (Salton McGill, 1983; Bookstein, 1985; Belkin Croft, 1987; Turtle Croft, 1990). Systems based on best match retrieval typically support simple natural language queries and automatic document indexing. This type of system has consistently performed much better than the exact match techniques under laboratory conditions using test collections of a few thousand records, and we ....

.... probabilistic retrieval model has been proposed which views information retrieval as an evidential reasoning process in which multiple sources of evidence about document and query content are combined to estimate the probability that a given document matches an information need (Turtle Croft, 1990). Different representations of the document content, different representations of the information need, and domain knowledge such as a thesaurus can all be taken into account under this model. INQUERY, a retrieval system based on this model, supports sophisticated indexing and complex query ....

[Article contains additional citation context not shown here]

Croft, W. B. & Das, R. (1990). Experiments with query acquisition and use in document retrieval systems. In Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368.

No context found.

Croft, W. Bruce, & Raj Das. 1990. Experiments with query acquisition and use in document retrieval systems. In Proceedings of the Thirteenth International ACM/SIGIR Conference, 349--365.

No context found.

W.B. Croft and R. Das. Experiments with Query Acquisition and Use in Document Retrieval Systems. In J. Vidick, editor, Proceedings of the 13th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 349--368. ACM Press, 1990.

No context found.

Croft, W.B, Das, R., "Experiments with Query Acquisition and Use in Document Retrieval Systems". in Proceedings of SIGIR 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC