Comparing Portuguese Opinion Lexicons in Feature-Based Sentiment Analysis

摘要

In this paper we evaluate different lexicons in feature level opinion mining on Brazilian Portuguese movie reviews. Research in this field often considers English data, while other languages are less explored. So we discuss and compare available resources and techniques that can be applied to Portuguese for dealing with this task. We found better results when using SentiLex adjetives. The results indicate a F-score of 0.73 for positive polarity recognition and 0.76 for negative polarity recognition.