Avaliação de critérios para seleção de sintagmas nominais com valor para a recuperação da informação

Abstract

This study assesses the criteria for selecting the most representative noun phrases from documents written in Portuguese in
the field of law. The research methods were literature review and an experiment. In the experiment, ten selection criteria were
applied to noun phrases extracted from a set of abstracts of theses and dissertations. The effectiveness of the criteria was
assessed regarding the selection of noun phrases relevant for information retrieval. Through the experiment, the most effective
criteria identified were removal of noun phrases with stopwords value or noun phrases containing pronouns, the selection
criteria of noun phrases based on position of occurrence, level of the noun phrase, inverse document frequency, and document
occurrence frequency.