Extração de Conhecimento em Texto Interpretando Opiniões, Percebendo Sentimentos

Abstract

 
Abstract. This paper presents a case study on Sentiment Analysis in a evaluative institutional process. In order to produce automatic analysis, it has been implemented a software named SentAnalysis. The software analyzes opinions written in the evaluative process, classifying them in positive, negative or neutral feelings. Results obtained from automatic evaluation were compared with those produced by two human evaluators. The comparison allowed us to recognize situations of partial and total convergence between system and evaluators. The evidences show that the developed system presents coherence with human evaluators, observing the fact that around 70% of the cases were of total convergence between human evaluators and system.