A lexicon pooled machine learning classifier for opinion mining from course feedbacks

Abstract

This paper presents our algorithmic design for a lexicon pooled approach for opinion mining from course feedbacks. The proposed method tries to incorporate lexicon knowledge into the machine learning classification process through a multinomial process. The algorithmic formulations have been evaluated on three datasets obtained from ratemyprofessor.com. The results have also been compared with standalone machine learning and lexicon based approaches. The experimental results show that the lexicon pooled approach obtains higher accuracy than both the standalone implementations. The paper, thus proposes and demonstrates how a lexicon pooled hybrid approach may be a preferred technique for opinion mining from course feedbacks and hence suitable for develpment in a practical caurse feedback mining system.