Computational Morphology for Italian

Any natural language processor, to be efficient and generalizable to unlimited or unrestricted linguistic domains, should try in our opinion, to simulate as closer as possible the human processor. Computational efficiency should not be obtained in spite of grammatical knowledge, but in accordance with linguistic models. It is rather counterintuitive to try to do away with grammatical knowledge in order to substitute it with heuristics which perform well, though in a limited domain. Linguistic processes as embodied by a grammar are organized into a set of modular components interacting with one another to produce their output. Our grammar is made out of the following components: a lexicon, a morphological analyzer, a syntactic preanalyzer coupled with a context-free grammar, and a parser