Machine learning for information retrieval: Neural networks, symbolic learning, and genetic algorithms

Abstract

Information retrieval using probabilistic techniques has attracted significant attention on the part of researchers in information and computer science over the past few decades. In the 198Os, knowledge-based techniques also made an impressive contribution to "intelligent" information retrieval and indexing. More recently, information science researchers have turned to other newer artificial-intelligence-based inductive learning techniques including neural networks, symbolic learning, and genetic algorithms. These newer techniques, which are grounded on diverse paradigms, have provided great opportunities for researchers to enhance the information processing and retrieval capabilities of current information storage and retrieval systems. In this article, we first provide an overview of these newer techniques and their use in information science research. To familiarize readers with these techniques, we present three popular methods: the connectionist Hopfield network; the symbolic ID3/ID5R; and evolution-based genetic algorithms. We discuss their knowledge representations and algorithms in the context of information retrieval. Sample implementation and testing results from our own research are also provided for each technique. We believe these techniques are promising in their ability to analyze user queries, identify users' information needs, and suggest alternatives for search. With proper user-system interactions, these methods can greatly complement the prevailing full-text, keywordbased, probabilistic, and knowledge-based techniques.

Genetic algorithms in relevance feedback: a second test and new contributions

“…ChenÕs group (Chen, 1995;Chen & Iyer, 1998) use a GA to learn the terms of a query that best represent a user-supplied set of documents (they call this process ''inductive query by example''). They follow with an intelligent agent that uses this algorithm to implement the feedback module.…”

“…The vectors corresponding to the documents provided as feedback are subjected to a conversion process such as that used by Chen (1995) to transform them into the chromosomes that our GA will work with. These chromosomes use a real representation, and will have the same number of genes (components) as the query and the feedback documents have terms with non-zero weights.…”

“…In recent years, there have appeared numerous applications of ''genetic algorithms'' (GAs) to information retrieval (Belew, 1989;Chen, 1995;Chen, Chung, & Ramsey, 1998;Chen & Iyer, 1998;Cordo on, Moya, & Zarco, 2000Gordon, 1988aGordon, ,b, 1991Horng & Yeh, 2000;Kraft, Petry, Buckles, & Sadasivan, 1994Lo opez-Pujalte, 2000;Lo opez-Pujalte, Guerrero Bote, & Moya Anego on, 2002a,b;Martı ın-Bautista, 2000;Martı ın-Bautista, Vila, & Larsen, 1999;Raghavan & Agarwal, 1987;Raghavan & Birchard, 1979;Robertson & Willett, 1994, 1996Sanchez, 1994;Sanchez, Miyano, & Brachet, 1995;Sanchez & Pierre, 1994;Smith & Smith, 1997;Vrajitoru, 1997Vrajitoru, , 1998Yang & Korfhage, 1992, 1993, 1994. Most of these applications refer to ''relevance feedback'', as is indicated in a review of the topic (Cordo on, Moya, & Zarco, 1999).…”

Local search: A guide for the information retrieval practitioner

“…There are some cases where local search techniques could be applied to any given task for a particular part of the taxonomy (see the discussion on stemming below). The practitioner must consider the issue of how a query (or profile in the case of some 'queryless' tasks) from a given task is represented (Chen, 1995). For example, weights may be best represented as a vector of floating-point numbers, and a term selection task as a binary string, one bit for each term to be included/excluded.…”

“…These problems can be NP hard and require the application of combinatorial optimisation methods, particularly the utilisation of local search. There has been a great deal of work in the area, and readers interested in reviews should refer to Chen (1995) and Sebastiani (2002). Having read the literature, we have identified a number of areas which need to be tackled in order to improve the quality of work in applying local search to IR problems:…”