Artigos científicos, teses e dissertações
Publicados nas melhores instituições de ensino

Proposition dun Système de Recherche dInformation Assistée par Ordinateur / Proposition of a Computer Aided Informations Retrieval Systems : with application in portuguese language.

Hélio Kuramoto

In this research paper, we propose a model to address problems typically faced by users of information indexing and retrieval systems. Two of the main issues are pointed out:low accuracy of information retrieval results; and difficulties typically encountered by novice users of the information retrieval system interface. Our research is limited to information retrieval systems (IRS) applied to full text databases. Through discussion of these problems we arrive at a solution that had been formerly proposed by the SYDO group: the use of nominal phrases (or Nominal Group) as descriptors instead of words which are generally used by the traditional IRS. In order to verify the feasibility of this proposition, we have developed a prototype of an IRS with a full text database. For this database, we chose a sample of scientific articles in the Portuguese language. Subsequently, we tested this prototype using a thesaurus, thereby enabling us to gain a more robust knowledge of : the behavior of nominal phrases in a tree structure ; and the behavior of the information retrieval interface. Our final step was to develop a model of recognition and excerpt of nominal phrases in Portuguese texts. In addition to the practical considerations of testing the feasibility of our proposition and developing a prototype, our work demonstrates that the knowledge obtained by the practice of excerpts and indexing of nominal phrases is important in establishing recognition and excerpts of nominal phrase models.