Opinion mining and emotion recognition in an intelligent learning environment

Abstract

In this paper, we present the development of an opinion-mining module. The development of the module consisted of creating an emotion tagged dataset of opinions; implementing an opinion mining module that processes sentences about computer programming, predicting or recognizing their polarity (positive/negative) and their type of emotion (frustrated, bored, excited, engagement, and neutral); and integrating the previous module in an intelligent learning environment. We evaluated the corpus, the accuracy of text polarity, and emotion recognition. The results with respect to polarity are promising (88.26%), however, the results in the detection of emotions are still low (60.0%). The reasons which likely explain these outcomes include a relatively small (7,777 records) and unbalanced corpus.