M. Gordon. Probabilistic and genetic algorithms in document retrieval. Communications of the ACM, 31(10):1208--1218, 1988.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....the specific document. However, in real applications there is hardly ever enough relevance information for a specific document available in order to estimate the required probabilities. For this reason, retrospective experiments based on this model (or related ones) might show its feasibility [18] [15], but are of little value with regard to real applications. The model described in [17] overcomes this This study was supported in part by the National Science Foundation under grant IRI 87 02735 problem by regarding document components as units to which the index term weights relate to; ....

M. Gordon. Probabilistic and genetic algorithms for document retrieval. Communications of the ACM, 31(10):1208--1218, 1988.

....determines the optimal one. Yang et al. 80] presented an evolutionary algorithm for query optimization by reweighting the document indexing without query expansion. Kraft et al. 81] apply genetic programming in order to improve weighted Boolean query formulation. Document Representation: Gordon [82] adopted a GA to derive better descriptions of documents. Here each document is assigned descriptions where each description is a set of indexing terms. Then genetic operators and relevance judgements are applied to these descriptions in order to determine the best one in terms of classification ....

M. D. Gordon, "Probabilistic and genetic algorithms for document retrieval, " Commun. ACM, vol. 31, no. 10, pp. 208--218, 1988.

....the optimal one. Yang et.al. 80] have presented an evolutionary algorithm for query optimization by re weighting the document indexing without query expansion. Kraft et.al. 81] apply genetic programming in order to improve weighted boolean query formulation. Document Representation: Gordon [82] adopted a GA to derive better descriptions of documents. Here each document is assigned N descriptions where each description is a set of indexing terms. Then genetic operators and relevance judgements are applied to these descriptions in order to determine the best one in terms of classification ....

M. D. Gordon, "Probabilistic and genetic algorithms for document retrieval," Communications of the ACM, vol. 31, no. 10, pp. 208--218, 1988.

....describing how close a solution guess is to the goal of the search, departing good from bad solutions. These two aspects represent the main di#culty of the GAs. Information retrieval researchers have suggested these algorithms to improve the performance of their systems. Gordon (1988 [8]) and Blair (1990 [1] have used them to improve document indexing. Chen (1995 [3] Petry et al. 1993 [14] Yang et al. 1992 [24] 1992 [24] Kraft et al. 1992 [11] and Sanchez et al. 1992 [19] present an approach based on GAs to enhance the query description. Finally, Gordon (1991 [9] ....

....the documents indexing (Salton 1971 [15] Vrajitoru 1997 [22] In our work, we have chosen to improve the document representation using a form of relevance feedback. To apply the GAs to this context, the genetic individuals must contain a representation of the whole collection. Gordon (1988 [8]) has applied GAs to a similar problem by improving the indexing of one document at a time. In this case, a genetic individual is a particular description of a document. If the collection is large, the cost of improving the document descriptions one by one can become too large. Considering this, ....

Gordon M. (1988) Probabilistic and Genetic Algorithms for Document Retrieval. Communications of the ACM. 31(10), 1208-1218.

....search spaces without having to evaluate all possible parameter values. An example of a heuristic method is the use of the Genetic Algorithm (GA) 68] a general optimization method based on genetic adaptation, to optimize how well the retrieval system is performing. Examples include Gordon s [56] application to adaptively determining the best document representations, and Yang at al. s [166] query optimization. The Genetic Algorithm modifies parameters of the documents or queries in order to optimize a measure of how well the retrieval system is performing. In both cases, the authors use ....

....be changed in order to improve system performance. Of course, gradient based optimization is not the only game in town. There are a variety of techniques available which can directly optimize a criterion without using gradient information. The Genetic Algorithm methods discussed in Chapter 2 [56] [166] exemplify such an approach, as do the downhill simplex method and direction set methods [103] Thus, average precision or other step like measures can be optimized directly. Press et al. 103, p. 291] does suggest that gradient information can often be very helpful in speeding up the ....

Michael Gordon. Probabilistic and genetic algorithms in document retrieval. Communications of the ACM, 31(10), October 1988.

....or absence of the key phrase in the document. The population of competing representations is modified over time by the genetic operators, with the goal being to evolve the set of key phrases which best describes the collection of documents making up a particular category. This is similar to [Gordon, 1988]. The genetic operators implemented for this model are crossover (initially single point) and mutation (single bit) The absence of one key phrase may be just as important as the presence of another in determining the classification of a document. For this reason the vector of key phrases ....

Gordon, M. Probabilistic and Genetic Algorithms for Document Retrieval. In Communications of the ACM, Vol. 31 No. 10.

....choices and behave differently even when applied repeatedly on the same data (Brasard and Bratley, 1994) Several researchers have used the GA in IR and their results seem to indicate that this algorithm could be efficient. In this vein, the main directions concern modifying the document indexing (Gordon, 1988; Blair, 1990) the clustering problem (Raghavan and Agarval, 1987; Gordon, 1991) and improving query formulation (Yang et al. 1992; Petry et al. 1993; Chen, 1995) In order to integrate the GA in our previous research (Vrajitoru, 1997) we have considered Gordon s model. One major problem ....

Gordon, M. (1988). Probabilistic and genetic algorithms for document retrieval. Communications of the ACM, 31(10), 1208-1218.

....of the terms. Bennett et al. [17] use a GA for optimising a query in a relational DBMS by minimising the cost of the query in terms of I O and CPU time. The queries used in this application were made up of relation and join pairs rather than terms and their weights. In the IR system used by Gordon [18], each document was given multiple descriptions with each description being a set of keywords that represent the subject of the document. The combination of terms in a descriptor were modified to improve retrieval performance by using user feedback on how the document should have been described. ....

....applications because the former use real parameter coding whereas the latter use artificial coding. The applications of GAs in IR have all followed the traditional low level chromosome representation even though they have not used an artificial coding scheme (e.g. binary coding) Gordon [18], Yang and Korfhage [16] and Jones et al. [15] use genetic operators that work on chromosomes of atomic genes, i.e. genes that cannot be subdivided, and use numbers as their genes. In contrast, Bennett at al [17] use genes that are not numbers and are composite, i.e. genes that can be subdivided; ....

[Article contains additional citation context not shown here]

Michael Gordon. Probabilistic and genetic algorithms for document retrieval. Communications of the ACM, 31(10):1208-- 1218, 1988.

.... the fact that there are many words that refer to the same concept [Streeter Lochbaum 1988] Gordon describes an approach more closely related to our approach, in which the organization of an information retrieval system evolves over time using a set of probabilistic genetic algorithms [Gordon 1988]. Connectionist Computing The idea of agents establishing graph edges in accordance with what resources exist is reminiscent of the learning notion introduced by connectionist computing ( neural network ) researchers: both cases involve an interconnection graph whose edges are somehow ....

M. Gordon. Probabilistic and Genetic Algorithms in Document Retrieval. Commun. ACM, 31(10), pp. 1208-1218, Oct. 1988.

....We have chosen not to explicitly optimize average precision because it is a discrete measure and is therefore not amenable to gradient based optimization. However, methods do exist for optimizing such non differentiable criteria, and some methods have been applied to tasks in Information Retrieval [7]. This may be a fruitful direction for further research. 3 Learning Weights on Phrases and Terms Our first application of the method involves combining two experts. This application demonstrates that the method can be used to evaluate expert performance in combination with other experts, in ....

Michael Gordon. Probabilistic and genetic algorithms in document retrieval. Communications of the ACM, 31(10), October 1988.

....time to accumulate some experience in its use 1 There have been several important efforts at developing a browser for news access ( 1] 4] although they are not based on the use of a user model. Most common approaches to information retrieval are based on the Boolean conjunction of keywords( 7] [13]) Recently ( 8] 10] this approach has been critically evaluated in a situation of broad vocabulary similar to our intended application. The alternative approaches of full text retrieval [12] and statistical information 1. All queries concerning the system should be referred to the first author ....

....other approaches to information retrieval that may be useful in the context of a personal news service. Most common by far are the approaches based on Boolean conjunction of keywords. There is a vast literature devoted to progressive improvement of the precision of the keyword based approach ( 7] [13]) Only recently ( 8] 10] has this approach been subject to rigorous evaluation in a setting of typical users. The Bellcore study [9] shows clearly that the limited vocabulary offered bykeyword based systems is entirely inadequate to deal with an information setting suitable for a very large set ....

Gordon. M. "Probabilistic and Genetic Algorithms for Document Retrieval" Communications of the ACM, 31(10),pp. 1208-1218, 1988

....[Ahamad et al. 1987] as opposed to instances of a class of objects, as in the current paper. Gordon developed a probabilistic scheme for supporting document retrieval, which focused on allowing the system to adaptively relate keywords over time, to improve document recall as the system was used [Gordon 1988]. Deering has worked on the problem of supporting multicast in an Internet environment [Deering 1988] Deering s work attempts to provide full delivery. In another paper we report simulation experiments for an implementation of SDMs based on modifications to Deering s multicast protocol [Angevine, ....

M. Gordon. Probabilistic and Genetic Algorithms in Document Retrieval. Commun. ACM, 31(10), pp. 12081218, Oct. 1988.

No context found.

Gordon, M. Probabilistic and genetic algorithms for document retrieval. Communications of the ACM, 31, 2 (1988), 152--169.

No context found.

M. Gordon, "Probabilistic and genetic algorithms for document retrieval," Communications of ACM, vol. 31, no. 2, pp. 152--169, 1988.

No context found.

M. Gordon. Probabilistic and genetic algorithms in document retrieval. Communications of the ACM, 31(10):1208--1218, 1988.

No context found.

Gordon, Michael (1988) Probabilistic and Genetic Algorithms for document retrieval. ACM 31:10, 1208-1218.

No context found.

Gordon M. 1988. Probabilistic and genetic algorithms for document retrieval. Communications of the ACNM, vol. 31, 1208-1218. 14

No context found.

Gordon, M.: Probabilistic and Genetic Algorithms for Document Retrieval. Communications of the ACM, 31 (10) (1988), 1208-1218.

No context found.

Michael Gordon. Probabilistic and genetic algorithms for document retrieval. Communications of the ACM, 31(10):1208--1218, October 1988.

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