D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....suggested by Paice [19] Based on recent publications on extended Boolean models, the P norm model is probably the most popular of the two well performing models. Greiff, Croft and Turtle [6] copied the behaviour of the P norm model in their inference network architecture and Losada and Barreiro [14] propose a belief revision operator that is equivalent to a P norm case. The vector space model and the probabilistic model stand for different approaches to information retrieval. The former is based on the similarity between query and document, the latter is based on the probability of ....

D.E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proceedings of the 22nd A CM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'99), pages 66-73, 1999.

....change (revision) must always lead to a consistent knowledge base. In addition, the revised knowledge base must contain the new information. Therefore, old knowledge maybe be deleted from the knowledge base, but this deletion should be minimal. The use of belief revision in IR was attempted in [61] as a way to compute the similarity of a document to a query for retrieval purpose. The Dalal s revision operator was chosen for implementing the belief revision process, since it provides an order among proposition interpretations, where propositions model index terms. The ordering is used to ....

D.E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean model. In Proceedings of ACM SIGIR, pages 66-73, Berkeley, CA, USA, 1999.

....suggested by Paice [19] Based on recent publications on extended Boolean models, the P norm model is probably the most popular of the two well performing models. Grei#, Croft and Turtle [6] copied the behaviour of the P norm model in their inference network architecture and Losada and Barreiro [14] propose a belief revision operator that is equivalent to a P norm case. The vector space model and the probabilistic model stand for di#erent approaches to information retrieval. The former is based on the similarity between query and document, the latter is based on the probability of ....

D.E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proceedings of the 22nd ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'99), pages 66--73, 1999.

....ranking function. To use possibilistic reasoning, we assume that there exists a similarity relation on the set D. The similarity relation can be imposed extraneously or generated automatically. One approach to the automatic generation of similarity relation is by using the Dalal s distance[2, 11]. Let A be Information Retrieval by Possibilistic Reasoning 5 a nite set and d 1 and d 2 be two documents, then the Dalal s distance between d 1 and d 2 is the proportion of A in which d 1 and d 2 do not agree, i.e. d 1 ; d 2 ) jfp 2 A : d 1 (p) 6= d 2 (p)gj jAj Thus, a similarity relation ....

....es , then they are ordered according to their nearness to since d ( corresponds to the minimal distance (or maximal similarity) from d to the documents satisfying . This can be seen as an interpretation of LUP in the possibilistic framework and in fact the same principle has been used in [11] in the case of Dalal s distance. However, we can further distinguish the documents satisfying by their distances to : that is 2. N d1 ( N d2 ( 0: thus d1 ( d2 ( 1, this means that both d 1 and d 2 have zero distance to since they satisfy by the re exivity of ....

D.E. Losada and A. Barreiro. \Using a belief revision operator for document ranking in extended boolean models". In Proceedings of the 22nd Annual International ACM SIGIR Conference in Research and Development of Information Retrieval, pages 66-73. ACM Press, 1999.

....suggested by Paice [19] Based on recent publications on extended Boolean models, the P norm model is probably the most popular of the two well performing models. Grei , Croft and Turtle [6] copied the behaviour of the P norm model in their inference network architecture and Losada and Barreiro [14] propose a belief revision operator that is equivalent to a P norm case. 2.4 Discussion The vector space model and the probabilistic model stand for di erent approaches to information retrieval. The former is based on the similarity between query and document, the latter is based on the ....

D.E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proceedings of the 22nd ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'99), pages 66-73, 1999.

No context found.

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.

No context found.

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. SIGIR-99, the 22nd ACM Conference on Research and Development in Information Retrieval, pages 6673, Berkeley, USA, August 1999.

No context found.

Losada, D. E. and Barreiro, A. (1999) Using a belief revision operator for document ranking in extended boolean models. In Hearst, M., Gey, F. and Tong, R. (eds), Proc. SIGIR99, 22nd ACM Conf. on Research and Development in Information Retrieval, Berkeley, USA, 15--19 August, pp. 66-- 73. ACM Press.

....details of the underlying model. Section 3 reports the experiments conducted and section 4 discusses the evaluation results and other relevant issues. The paper ends with some conclusions. 2 Background In this work we follow the logical approach for IR suggested by Losada and Barreiro [11, 15, 10]. This model is based on the combined use of Propositional Logic and Belief Revision. Along this paper, we will refer to this model as PLBR model. There are a number of reasons supporting this election. First, the PLBR model was e ciently implemented and polynomial time algorithms were supplied to ....

....into false . As a consequence, the application of the logical entailment to decide relevance would assign the same status to both d 1 and d 2 with respect to the query q. This is not appropriate for IR purposes because d 1 is likely more relevant than d 2 (d 1 ful lls partially the query) In [11] a method to get a non binary measure of the entailment d j= q was proposed. To de ne a non binary measure of relevance the distance from each model of d to the set of models of q is measured. In the eld of Belief Revision (BR) measures of distance between logical interpretations are formally ....

[Article contains additional citation context not shown here]

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. SIGIR-99, the 22nd ACM Conference on Research and Development in Information Retrieval, pages 6673, Berkeley, USA, August 1999.

....semantics that should be used for changing the retrieval situation with the document. Finally, as logical entailment is too strict to decide relevance, we propose to use distances between logical interpretations to measure the uncertainty of (S ffi d) j= q, following the techniques presented in [14] which allows us to rank relevant documents. The rest of this paper is organized as follows. Section 2 explores the connection between conditionals and change semantics, which is formalized by the Ramsey test. That section also discusses Grdenfors Triviality result. Section 3 analyzes the kind ....

....changed situation will be those models of the document having less keywords in disagreement with respect to the models of the situation. Clearly, this behaviour is related to IR, where there are several similarity measures counting term matches between representations. In this line, a recent work [14] presents the use of Dalal s operator to get a measure of the uncertainty of d j= q. Our use of Dalal s operator is different here because we are interested in the final result of the revision. If the revised situation entails the query, the document is considered relevant. In contrast, the key ....

[Article contains additional citation context not shown here]

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. of SIGIR-99, the 22th ACM Conference on Research and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.

....of the model against some IR test collections. Next, we extend the model to deal with relevance feedback. The evaluation results of the model with feedback are presented in section 5. The paper ends with some conclusions. 2. The PLBR Model Along this work we will use the model proposed in [11]. This section depicts the basic foundations of this model. The review is intentionally brief and we refer to [11] for a detailed description of the model. 2.1. Representation and matching Documents and queries are represented as Propositional Logic formulas. The expressiveness of propositional ....

....evaluation results of the model with feedback are presented in section 5. The paper ends with some conclusions. 2. The PLBR Model Along this work we will use the model proposed in [11] This section depicts the basic foundations of this model. The review is intentionally brief and we refer to [11] for a detailed description of the model. 2.1. Representation and matching Documents and queries are represented as Propositional Logic formulas. The expressiveness of propositional formulas allows us to manage representations which are the logical counterpart of binary weighted vectors, e.g. d ....

[Article contains additional citation context not shown here]

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. of SIGIR-99, the 22th ACM Conference on Research and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.

.... logic allows us to model binary weighted vectors, e.g. d = information science :maths, but more expressive representations can also be handled, e.g. d = relevance feedback) document f iltering) In order to measure the relevance of a document d to a query q, we use the method proposed in [4] to get a non binary measure of the entailment d j= q. An important circumstance is that this model was eciently implemented [5] and, furthermore, evaluation against small collections was made [6] We focus on a feedback process based on selecting terms from retrieved documents. Term selection ....

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. ACM SIGIR-99, pages 66-73, Berkeley, USA, August 1999.

....of data require a minimum response time and, therefore, very efficient procedures. As a result, logical models of IR have to take great care of their complexity results. In this work we implement document ranking within the Belief Revision (BR) framework. We follow the theoretical development of [13], where a ranking of documents given a query is obtained within Dalal s revision [5] A document and a query are represented in Propositional Logic by logical formulas d and q respectively and Belief Revision procedures are suitable for quantifying the uncertainty of d q. This research goes in ....

....and Belief Revision procedures are suitable for quantifying the uncertainty of d q. This research goes in the line of some works [4, 15, 2, 18, 14, 8] that have followed different techniques in order to implement Van Rijsbergen s uncertainty principle [17] However the model proposed in [13] needs to handle logical interpretations and, thus, its implementation would take exponential time to decide relevance. To overcome this problem, we propose a syntactic characterization for the formulas representing documents and queries. This way, we are able to develop efficient algorithms ....

[Article contains additional citation context not shown here]

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proc. of SIGIR-99, the 22th ACM Conference on Research and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.

.... as a theory and a document representation d as a new information, the BR process q ffi d produces a measure of the uncertainty of d q, in the line proposed by van Rijsbergen in [7] The application of this BR process to the field of Information Retrieval (IR) has been introduced in a recent work [6]. The BR operator suggested by Dalal [3] ffi D , was proposed to carry out the revision, q ffi D d. An important result in this direction is that using Dalal s operator ensures theoretically an ordering among the documents which follows the notion of proximity to the query. In [6] the BR ....

....a recent work [6] The BR operator suggested by Dalal [3] ffi D , was proposed to carry out the revision, q ffi D d. An important result in this direction is that using Dalal s operator ensures theoretically an ordering among the documents which follows the notion of proximity to the query. In [6] the BR framework was used to build a similarity measure, BRsim, that was shown equivalent to the inner product matching function, which is frequently used in the vector space model. However, this equivalence was done restricting the expressiveness of the query. This decision was made in order to ....

D. E. Losada and A. Barreiro. Using a belief revision operator for document ranking in extended boolean models. In Proceedings of SIGIR-99, the 22th ACM Conference on Research and Development in Information Retrieval, pages 66--73, Berkeley, California, August 1999.

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