Extração automática de definições : um estudo de caso em textos legislativos

Abstract

This work presents a variation of an automated technique of natural language processing of Brazilian Portuguese texts intended for building a glossary from domain specific texts that contain definitions amongst its content. The results of a case study on applying the technique to a corpus of texts on telecommunications regulations are shown. The usual chain of natural language processing has been followed: preprocessing, segmentation, and partofspeech tagging. A set of feature extraction functions is then specified for extracting definitions. The associated weights are then determined by applying a machine learning technique on a training corpus, which is part of the reference glossary. At lest, the definitions are extracted from the texts and evaluated upon a test corpus, which is also part of the reference glossary. The results are then discussed in light of other definition extraction techniques.