An approach based on techniques of state-finite has applied to the processes of unification of terms in Spanish. The algorithms of conflation are computational procedures utilized in some Information Retrieval (RI) systems for the unification of term variants, semantically equivalent, to a normalized form. The programs that carry out habitually this process are called: stemmers and lematizadores. The objective of this work is to evaluate the deficiencies and errors of the lematizadores in the conflation of terms. The method utilized for the construction of the lematizador has been based on the implementation of a linguistic tool that permits to build electronic dictionaries represented internally in Finite-State Transducers (FST). The lexical resources developed have applied to a corpus of verification to evaluate the performance of these lexical parsers. The metric of evaluation utilized has been an adaptation of coverage and precision measures. The results show that the main limitation of unification processes of tem variants through technology of state-finite is the under-analysis.
Term conflation; Lemmatization; Finite-state transducers
