Enhancing sentiment analysis using syntactic patterns

Abstract

Using specialized analysis tools, combining natural language processing techniques with machine-learning-based sentiment analysis, it is possible to establish positive and negative sentiments expressed in opinion texts. Thus, organizations have the possibility to act in an adequate way, having the opportunity to improve their relationship with their customers and improve their loyalty, according to the type of sentiment identified. In this paper we present and describe a sentiment analysis system especially developed to identify sentiments, of different polarities, expressed in opinion texts of students of an eLearning application. We have slightly rewritten the usual way of approaching sentiment analysis problems by using Hearst patterns, for improving classification models efficiency, valuing the sentiments expressed in a wider scale of classification values.