Um ambiente para a analise superficial de linguas baseado em automatos finitos

AUTOR(ES)
DATA DE PUBLICAÇÃO

1998

RESUMO

Syntactic analysis is an important component of most natural language processing systems. Typically parsers were implemented using techniques derived from Context Free Grammars. The increasing need for efficient systems was one of the reasons to search for new approaches to syntactic analysis. Some natural language applications do not need complete parsing of the deep structure of the sentences (derivation trees). In these cases, a representation of the surface relations among words in a text is enough. The goal of a surface parser is to recognize the natural language sentences providing information only about some particular features. It is not concerned with the deep structure of the sentences. This kind of parsers are more efficient and can replace a parser implementing a complete syntactic analysis in different situations. We propose in this dissertation an environment to implement surface parsing of natural languages. In our approach every sentence is represented by a sequence of part-of-speech tags caled syntactic pattern. The hypothesis underlying our work is that the core of the sentences used in natural language texts can be captured with a computationally tractable number of syntactic patterns. Data structures based on finite-state automata have been used in representing1large word vocabularies. Syntactic patterns are strings of symbols that can be compared in some sense with words. We have shown that finite-state automata are adequate to store syntactic patterns. Furthermore they allow an efficient implementation of the recognizing algorithms, and other more complex ones, as the agreement adviser presented in this work. One of the problems common to many approaches for syntactic parsing is the lack of a method or information source to build a system capable of mastering a real example. As an option, we propose the collection of syntactic patterns from annotated corpora of texts

ASSUNTO(S)

processamento da linguagem natural (computação) algoritmos

ACESSO AO ARTIGO

Documentos Relacionados