Acceso usuarios
José Vega, Antoine Ogonowski
This paper describes an information retrieval system called NLA which analyses natural language queries. The languages used here are: English, Italian & French. The query analysis is performed in two stages: the comprehension stage, where the user's request is translated into an unambiguous internal form using linguistic analysers, and the search stage where a documentary machine uses the internal representation in order to map it into facts (predicates) recorded in a knowledge base. The documentary machine then tries to broaden the relevant concepts (Normalised Terms [NT]) of the query in order to find other NT's semantically close to the same domain. Finally, replies are worked out by calculating answers to each constraint (predicate-NT pair) through applicative rules and cross-referencing search strategies using Boolean equations. This system is also used to classify and index the records (files) in the Dianeguide database. The internal representation of indexed records is a dependency tree where the leaves are predicates similar to those produced by the natural language interface.
