A. S. Mandar Mitra and C. Buckley. Improving automatic query expansion. In Proceedings of ACM SIGIR 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for the word swing , it will be signed as relevant. This document is irrelevant, in fact, because the main topic of the document is about the swing function and its properties in Java rather than the Java programming language. Although adding new words to queries enhances their performance [11], most attempts at automatically expanding queries have failed to improve retrieval effectiveness [12] However, Qiu [13] suggests that this is because two problems in were not solved: the selection of suitable terms and the weighting of selected additional search terms. Furthermore, we see the ....

Mitra M., Singhal A., and Buckley C. `Improving automatic query expansion'. W.B. Croft, A., C.J. van Rijsbergen, R. Wilkinson, and J. Zobel, editors, Proceedings of the 21st Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, pp. 206--214, August 1998.

....lexically related terms ( 24, 16] A drawback of this approach is that a lexical reference cannot capture idiosyncrasies of a corpus. Prior results of using lexical references for query expansion are not encouraging. The third class of techniques use relevance feedback to expand a query ([18, 25]) Relevance feedback is e#ective only if accurate relevance feedback information is available, which requires user intervention. Most traditional work on comparing term distributions in two corpora applies statistical tests, such as # test, to compare the distributions of terms in the corpora ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR1998.

....That is, while current P2P systems support naive text match search, they cannot support semantic search. As a result, the systems can only find data items which are given a keyword (or meta data) exactly indicated in a query. The most familiar mechanism enabling semantic search is query expansion [16, 17], which has been investigated as an IR technique for several decades. Query expansion means adding relevant terms to the original query. The purpose of query expansion is to cope with the mismatch between the term used by searcher and that expected by writers of the documents. This mismatch may be ....

Mandar Mitra, Amir Singhal, and Chris Buckley. Improving Automatic Query Expansion. Proc. ACM SIGIR '98, Aug. 1998.

....A simple example is a user typing in Mchae Jordan. If the user is looking for sports related results, a better query might be Michael Jordan and basketball, helping to reduce the chance of a document being returned about the country of Jordan, or a different Michael Jordan. Mitra et al. [15] describe an automatic approach to discover extra query terms that can improve search precision. Their basic algorithm, like other relevance feedback algorithms, retrieves an initial set of possibly relevant documents, and discovers correlated features to be used to expand the query. Unlike other ....

....entropy loss to provide an approximation of the usefulness of the individual feature. This approach will correctly assign low scores to features that, although common in both sets, are unlikely to be useful for a binary classifier. 3. 2: Choosing Query Modifications Like the work of Mitra [15], the goal of our query modification is to identify features that could enhance the precision of a query. Unlike their work, we have extra information regarding the user s intent in the form of labelled data. The labelled data de fines a category, and the learned modifications can be re applied ....

Mandar Mitra, Amir Singhal, and Chris Buckley. Improving automatic query expansion. In ACM SIGIR 98, Melbourne Australia, 1998. ACM.

....reasonable, that is likely to produce superior queries at least occasionally) paraphrases. We should note here, that traditionally, the idea of adding words to a query is called query expansion and is done either manually or automatically (in which case it is called pseudo relevance feedback [18]. We have chosen a basis composed of 15 operators, grouped into four categories: DELETE, REPLACE, DISJUNCT, and OTHER. We use five DELETE operators (delete all prepositions, delete all wh words, delete all articles, delete all auxiliaries, and delete all other stop words based on a list of 163 ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proc. SIGIR98, Melbourne (AU), 1998.

....develop a query relaxation framework for searching answers that match the given query conditions approximately . Query relaxation enables systems to relax the user query to a less restricted form to permit approximate answers. Query relaxation technique has been used in relational databases (e.g. [MSB98, CCH94, CYC96, CG99, Gaa97]) and has proven to be a valuable technique for deriving approximate answers. nature of XML model allows varied structures and or values and the non rigid XML tag syntax enables to embed a wealth of meta information in XML documents. The following points illustrates that query relaxation becomes ....

M. Mitra, A. Singhal, and C. Buckley. "Improving Automatic Query Expansion". In ACM SIGIR, Melbourne, Austrailia, Aug. 1998.

....of the crawling algorithms and gauge their performance against their e#ciency. 4 Characterization of Topics The third dimension of our evaluation framework pertains to topic characteristics. In information retrieval research it is understood that query characteristics a#ect performance [30, 36, 3, 28]. In the classic 1988 study by Saracevic and Kantor [36] query characteristics were explored within a larger context that included the study of users and search methods. Their questions were classified by expert judges regarding: domain (subject) clarity, specificity, complexity and ....

....high complexity and many presuppositions. Beaulieu et al. correlated search outcomes with query characteristics examining aspects such as topic type [3] Mitra and colleagues explore the e#ect of query expansion strategies by di#erentiating queries based on their initial retrieval performance [28]. There is also active research on the types of queries users input to search engines [38, 19] For example in [38] the authors study over a million queries posed to the Excite search engine and find that the language of Web queries is distinctive in that a great many terms are unique. A key ....

M Mitra, A Singhal, and C Buckley. Improving automatic query expansion. In Proc. 21st ACM SIGIR Conf. on Research and Development in Information Retrieval, pages 206--214, 1998.

....element of relevance feedback can be omitted: since, statistically, a reasonable proportion of highly ranked documents are relevant, these can be used to expand the query without interaction with the user. Experiments have consistently shown that query expansion leads to improved performance [1, 18, 36]. In practice, however, query expansion has a significant disadvantage: it greatly increases the number of query terms, thus increasing the costs of query evaluation. A query evaluation technique that was e#cient for small queries only such as the two word to five word queries typically ....

....dominant as query length grows. These results confirm our supposition that, as query length increases, DO processing becomes relatively more expensive. For long queries TO or TOS evaluation is clearly preferable, as it is for use with techniques such as relevance feedback or query expansion [1, 18, 36], which greatly increase the number of query terms. To confirm our observations we built a document collection of short pieces of text of 50 500 bytes each, from the same data; these pieces of text were identified by using paragraph breaks where possible, but introducing breaks or coalescing ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In W.B. Croft, A. Mo#at, C.J. van Rijsbergen, R. Wilkinson, and J. Zobel, editors, Proceedings of the 21st Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, pages 206--214, August 1998.

....reasonable, that is likely to produce superior queries at least occasionally) paraphrases. We should note here, that traditionally, the idea of adding words to a query is called query expansion and is done either manually or automatically (in which case it is called pseudo relevance feedback [18]. We have chosen a basis composed of 15 operators, grouped into four categories: DELETE, REPLACE, DISJUNCT, and OTHER. We use five DELETE operators (delete all prepositions, delete all wh words, delete all articles, delete all auxiliaries, and delete all other stop words based on a list of 163 ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proc. SIGIR98, Melbourne (AU), 1998.

....search concepts, not just one. The researchers point out that concept scoring improves the ranking of documents that contain representatives of all search concepts. The problem of focusing the query (or query drift) in QE based on relevance feedback has been discussed by Mitra, Singhal and Buckley [38]. The researchers state that all aspects of a request should be represented in documents assumed to be relevant and used as a source for expansion keys. These studies corroborate our results of the effectiveness of concept based query structures combined with QE. However, we went further by ....

Mitra, M., Singhal, A. and Buckley, C. Improving automatic query expansion. In: Croft, W.B., Moffat, A, van Rijsbergen, C. J., Wilkinson, R. and Zobel, J. eds. Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval. New York, NY: ACM, 1998, 206--214.

....to the topics of Figure 1 confirm the intuition given in Section 1 about the generality of the two topics and its e#ect on the relative performance of the two crawlers. 5. RELATED RESEARCH In information retrieval research it is understood that query characteristics a#ect performance[19, 22, 2, 18]. In the classic 1988 study by Saracevic and Kantor [22] query characteristics were explored within a larger context that included the study of users and search methods. They found for example that the number of relevant documents retrieved was higher in questions of low clarity, low specificity, ....

....high complexity and many presuppositions. Beaulieu et al. correlated search outcomes with query characteristics examining aspects such as topic type [2] Mitra and colleagues explore the e#ect of query expansion strategies by di#erentiating queries based on their initial retrieval performance [18]. There is also active research on the types of queries users input to search engines [23, 13] For example in [23] the authors study over a million queries posed to the Excite search engine and find that the language of Web queries is distinctive in that a great many terms are unique. A key ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proc. 21st ACM SIGIR Conf. on Research and Development in Information Retrieval, pages 206--214, 1998.

....relationships in an analysis system, rather it is a requirement. The problem the vector space model (largely used in information retrieval) is faced with is notably the lack of these considerations. Many researchers have limited the use of lexical devices on the level of query expansion [12]. Query expansion is the set of techniques for modifying a query in order to satisfy an information need. Basically, terms are added to an existing query (focus on short queries) causing a modification of terms weights. In our approach, we go one step further in integrating lexical knowledge as ....

M. Mitra, A. Singhal, and C. Buckley. Improving Automatic Query Expansion. In Proc. of the 21st Annual Int. ACM SIGIR Conf. on Research and Development in Information Retrieval, pp 206--214, Melbourne, August 24 - 28 1998.

....with related terms. Term weights were recomputed by using the standard Rocchio method [15] where we considered the top 10 documents to be relevant and the bottom 250 documents to be non relevant. We allowed at most 20 terms to be added to the original query. We did not carry out any filtering [11] before applying Rocchio, since some experiments that we carried out on the CLEF 2000 data set indicated a decrease in retrieval effectiveness. 2.2 Inflectional Morphology Previous retrieval experimentation [6] in English did not show consistent significant improvements by applying morphological ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 206--214, 1998.

....with related terms. Term weights were recomputed by using the standard Rocchio method [10] where we considered the top 10 documents to be relevant and the bottom 250 documents to be non relevant. We allowed at most 20 terms to be added to the original query. We did not carry out any filtering [7] before applying Rocchio, since some experiments that we carried out on the CLEF 2000 data set indicated a decrease in retrieval effectiveness. 1 2.2 Inflectional Morphology Previous retrieval experimentation [3] in English did not show consistent significant improvements by applying ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 206--214, 1998.

....the MT system, but additionally used the German English bilingual wordlist. The two German monolingual runs were a simple, straightforward retrieval run, and a run that was enhanced through blind relevance feedback (for a discussion of blind feedback and some possible enhancements to it, see e.g. Mitra et al. 1998) The choice of relevance feedback was to imitate the expansion effect of the similarity thesaurus for the other languages. We expanded the query by the twenty statistically best terms from the top 10 initially retrieved documents. The two runs per each language are merged by adding together the ....

M. Mitra, A. Singhal, and C. Buckley. Improving Automatic Query Expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 206 - 214, 1998.

....bootstrapping technique. Schiffman and McKeown [19] describe experiments in automatically building a lexicon of phrases from a collection of documents with the goal of building an index of the collection that is better suited for question answering. Also related is a large body of research (e.g. [12]) that describes methods for automatically expanding queries based on the relevance of terms in the top ranked documents. An interesting approach presented in [23] describes how to automatically expand a query based on the co occurrence of terms in the query with the terms in the top ranked ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In ACM SIGIR, 1998.

....bootstrapping technique. Schiffman and McKeown [19] describe experiments in automatically building a lexicon of phrases from a collection of documents with the goal of building an index of the collection that is better suited for question answering. Also related is a large body of research (e.g. [12]) that describes methods for automatically expanding queries based on the relevance of terms in the top ranked documents. An interesting approach presented in [23] describes how to automatically expand a query based on the co occurrence of terms in the query with the terms in the top ranked ....

M. Mitra, A. Singhal, and C. Buckley. Improving automatic query expansion. In ACM SIGIR, 1998.

....effectiveness are the main focus of information retrieval (IR) researchers. The two measures used to evaluate IR algorithms are efficiency and effectiveness. There are many ways to improve accuracy, two of which are Proximity Search [Goldman et al., 1998] and Relevance Feedback [Rocchio, 1971, Mitra et al., 1998, Chang et al., 1999] Proximity search improves the result of the query processing by creating the capability to search for phrases, or terms in a specific window size. An example is the implementation of the Proximity Search on the King James Bible by the University of Michigan. The system ....

M. Mitra, A. Singhal and C. Buckley, Improving Automatic Query Expansion, ACM SIGIR'98.

....A simple example is a user typing in Michael Jordan. If the user is looking for sports related results, a better query might be Michael Jordan and basketball, helping to reduce the chance of a document being returned about the country of Jordan, or a different Michael Jordan. Mitra et al. [15] describe an automatic approach to discover extra query terms that can improve search precision. Their basic algorithm, like other relevance feedback algorithms, retrieves an initial set of possibly relevant documents, and discovers correlated features to be used to expand the query. Unlike other ....

....entropy loss to provide an approximation of the usefulness of the individual feature. This approach will correctly assign low scores to features that, although common in both sets, are unlikely to be useful for a binary classifier. 3. 2: Choosing Query Modifications Like the work of Mitra [15], the goal of our query modification is to identify features that could enhance the precision of a query. Unlike their work, we have extra information regarding the user s intent in the form of labelled data. The labelled data defines a category, and the learned modifications can be re applied for ....

Mandar Mitra, Amit Singhal, and Chris Buckley. Improving automatic query expansion. In ACM SIGIR 98, Melbourne Australia, 1998. ACM.

....Through the replacement step, we obtain a possibly shorter French result list. We then use the top documents from this list to do a pseudo relevance feedback loop, i.e. we select the most significant terms from this set of documents using methods developed for relevance feedback see also [3] and [5]) These terms form our French query, which we run against the French documents, to obtain a French result list. Note that the multilingual collections used for document alignment do not necessarily have to be identical to the search collections, although they were in the case of our TREC ....

M. Mitra, A. Singhal and C. Buckley. Improving Automatic Query Expansion. In Proceedings of the 21 st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 206-214, 1998.

....A simple example is a user typing in Michael Jordan. If the user is looking for sports related results, a better query might be Michael Jordan and basketball, helping to reduce the chance of a document being returned about the country of Jordan, or a different Michael Jordan. Mitra et al. [15] describe an automatic approach to discover extra query terms that can improve search precision. Their basic algorithm, like other relevance feedback algorithms, retrieves an initial set of possibly relevant documents, and discovers correlated features to be used to expand the query. Unlike other ....

....sorted by expected entropy loss to provide an approximation of the usefulness of the individual feature. This approach assigns low scores to features that, although common in both sets, are unlikely to be useful for a binary classifier. 3. 2: Choosing Query Modifications Like the work of Mitra [15], the goal of our query modification is to identify features that could enhance the precision of a query. Unlike their work, we have extra information regarding the user s intent in the form of labelled data. The labelled data defines a category, and the learned modifications can be re applied for ....

Mandar Mitra, Amit Singhal, and Chris Buckley. Improving automatic query expansion. In ACM SIGIR 98, Melbourne Australia, 1998. ACM.

....and hence, query modification. When a query modification is requested by the niche search engine, the query modification closest to the desired operating point is used. Practical deployment of query modifiers is a major area of investigation by itself, which we will not explore in detail (see [20, 6, 19, 21] for a discussion) However, one major problem is that secondary search engines impose limits on operators and query lengths. We now address this problem in our framework. In off line applications, procedures such as the QuineMcCluskey method, or approximations thereof, can be used to simplify ....

M. Mitra, A. Singhal, and C. Buckley, "Improving automatic query expansion," in Proc. SIGIR98, (Melbourne (AU)), ACM, 1998.

....a modified Rocchio approach with the additional filter of requiring the term to occur in at least 2 of the top documents (N 1. These efforts increased our average precision recall to 0.2359. We note that similar work has been done earlier (most notably Mitra, Singhal and Buckley, SIGIR 1998 [Mitra98] but our specific variations (automatic title concepts expanded with k stems and N 1) are new and effective. When we ran the second pass, we loosened the restriction of requiring at least one word from ALL title concepts to requiring at least one word from ONE of the concepts. In order to ....

Mitra, M., A. Singhal, C. Buckley. "Improving Automatic Query Expansion". Proceedings of the Twentyfirst Annual International ACM SIGIR Conference on Research and Development in Information Retrieval ACM SIGIR'98 pages 206-214, 1998.

No context found.

A. S. Mandar Mitra and C. Buckley. Improving automatic query expansion. In Proceedings of ACM SIGIR 1998.

No context found.

M. Mitra, A. Singhal and C. Buckley. Improving Automatic Query Expansion. Proceedings of the ACM International Conference on Research and Development in Information Retrieval (SIGIR), Melbourne, Australia, 1998.

No context found.

Mitra, M., A. Singhal, C. Buckley. "Improving Automatic Query Expansion". Proceedings of the TwentyFirst Annual International ACM SIGIR Conference on Research and Development in Information Retrieval ACM SIGIR'98 pages 206-214, 1998.

No context found.

Mitra, M., Singhal, A., and Buckley, C. (1998). Improving Automatic Query Expansion. In SIGIR '98 (pp. 206--214).

No context found.

Mitra, M., Singhal, A., and Buckley, C. (1998). Improving automatic query expansion. In Research and Development in Information Retrieval, pages 206--214.

No context found.

A. S. Mandar Mitra and C. Buckley. Improving automatic query expansion. In Proceedings of ACM SIGIR 1998.

No context found.

Mandar Mitra, Amit Singhal, and Chris Buckley. Improving automatic query expansion. In Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval, pages 206--214. ACM Press, 1998.

No context found.

M. Mitra, A. Singhal, C.Buckley, Improving Automatic Query Expansion, SIGIR '98, 206-214, 1998

No context found.

Mitra, M., Singhal A., and Buckley, C. Improving Automatic Query Expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference. ACM, 1998.

No context found.

Mitra, M., Singhal, A. and Buckley, C., Improving Automatic Query Expansion. In Proceedings of the ACM SIGIR 1998.

No context found.

M. Mitra, A. Singhal, and C. Buckley, "Improving Automatic Query Expansion," Proc. 21st Ann. Int'l ACM SIGIR Conf. Research and Development in Information Retrieval, pp. 206-214, 1998.

No context found.

Mandar Mitra, Amit Singhal, and Chris Buckley. Improving automatic query expansion. In Research and Development in Information Retrieval, pages 206--214, 1998.

No context found.

Mitra, M., Singhal, A. and Buckley, C. (1998) Improving Automatic Query Expansion. In Proc. of the 21 st ACM SIGIR Conference, pp. 206 - 214.

No context found.

Mitra, M., Singhal, A., and Buckley, C., "Improving automatic query expansion," Proc. 21 Annual ACM SIGIR Conf. On Research and Development in Information Retrieval, 206-214, 1998. Available at http://citeseer.nj.nec.com/121460.html

No context found.

M. Mitra, A. Singhal and C. Buckley, "Improving Automatic Query Expansion," In Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval (pp. 206 - 214). Melbourne, Australia: ACM Press.

No context found.

M. Mitra, A. Singhal, and C. Buckley, "Improving Automatic Query Expansion," Proc. 21st Ann. Int'l ACM SIGIR Conf. Research and Development in Information Retrieval, pp. 206-214, 1998.

No context found.

M. Mitra, A. Singhal, and C. Buckley. Improving Automatic Query Expansion. In SIGIR'98, 1998.

No context found.

Mitra. M., Singhal, A., and Buckley, C.: Improving Automatic Query Expansion. In: Proc. of the 21st Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval (1998) 206-214

No context found.

Mitra, M., Singhal, A. and Buckley, C. (1998) Improving Automatic Query Expansion. In Proc. of the 21 st ACM SIGIR Conference, pp. 206 - 214.

No context found.

M. Mitra, A. Singhal, C.Buckley, Improving Automatic Query Expansion, SIGIR '98, 206-214, 1998.

No context found.

A. S. Mandar Mitra and C. Buckley. Improving automatic query expansion. In Proceedings of ACM SIGIR 1998.

No context found.

M. Mitra, A. Singhal, C. Buckley. Improving Automatic Query Expansion. Proceedings of SIGIR '98, August 1998, pp. 206-214. 69

No context found.

Mitra, M., Singhal, A., and Buckley, C. Improving Automatic Query Expansion. In Proceedings of the 21 st Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, 206214.

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