LEARNING SYNTACTIC RULES AND TAGS WITH GENETIC ALGORITHMS FOR INFORMATION-RETRIEVAL AND FILTERING - AN EMPIRICAL-BASIS FOR GRAMMATICAL RULES

The grammars of natural languages may be learned by using genetic algo rithms that reproduce and mutate grammatical rules and part-of-speech tags, improving the quality of later generations of grammatical compon ents. Syntactic rules are randomly generated and then evolve; those ru les resulting in improved parsing and occasionally improved retrieval and filtering performance are allowed to further propagate. The LUST s ystem learns the characteristics of the language or sublanguage used i n document abstracts by learning from the document rankings obtained f rom the parsed abstracts. Unlike the application of traditional lingui stic rules to retrieval and filtering applications, LUST develops gram matical structures and tags without the prior imposition of some commo n grammatical assumptions (e.g. part-of-speech assumptions), producing grammars that are empirically based and are optimized for this partic ular application.