Reconhecimento automático de expressões temporais

Abstract

In this work, we explore the task of identifying and classifying temporal expressions in natural language text. We present an overview of the areas of Natural Language Processing, Information Extraction, and Named Entity Recognition, and show the difficulties inherited to the task of automatically recognizing expressions involving temporal elements. By using Machine Learning methods with a supervised learning approach, we propose a set of models that try to locate temporal expressions and correctly classify them into predefined categories. The best model achieves 74% of f-measure in our corpus, which is par with the state of the art for Portuguese. However, it is the only approach we know of that does not rely on manually created rules or regular expressions.