Security in software engineering requirement

Abstract

As the most extensively studied topic in sentiment analysis, sentiment classification has mainly two types of methods: supervised learning and unsupervised learning. As one of unsupervised learning methods, sentiment lexicon-based method plays a very important role in sentiment classification. However, there are two problems in existing sentiment lexicon-based method. Firstly, Sentiment words are only divided into positive and negative categories in existing sentiment lexicons, but polarity intensity of sentiment words is not quantified. Secondly, sentiment classification is formulated as an either-or problem, yet the fuzziness of sentiment categories is not considered. In order to solve the two problems, we propose a new method based on quantified sentiment lexicon and fuzzy set. We firstly construct some quantified sentiment lexicons based on three Chinese sentiment lexicons, and then calculate sentiment intensity of Chinese reviews by quantified sentiment lexicon, finally, we classify Chinese reviews based on fuzzy classifier. Experiment results in two review datasets demonstrate that our method outperforms the state-of-the-art methods.