Análisis morfológico automático del español a través de generación
Alexander F. Gelbukh, Grigori Sidorov, Francisco Velásquez · Escritos. Revista del Centro de Ciencias del Lenguaje · 2003
Most widely spread method used in automatic morphological analysis systems is based on the well-known twolevel morphology model. However, this model is not well suit for languages with irregular stem alternations (e.g., Spanish or Russian). In this work we describe a system with automatic morphological analysis for Spanish based on another model, based on analysis through generation. This model consists of a set of rules that allow for obtaining of all possible stems for each lexeme, storing them in the dictionary, producing all possible hypotheses during analysis, and verification of the hypotheses through morphological generation. We used a dictionary containing 40,000 lemmas, with which more than 2.500.000 possible grammatical forms can be recognized. For treatment of unknown words, a heuristic-based algorithm is being developed. The system is freely available for academic use.