Sentiment analysis of text using deep convolution neural networks

Abstract

Sentiment analysis has been one of the most researched topics in Machine learning. The roots of sentiment analysis are in studies on public opinion analysis at the start of 20th century, but the outbreak of computer-based sentiment analysis only occurred with the availability of subjective text in Web. The task of generating effective sentence model that captures both syntactic and semantic relations has been the primary goal to make better sentiment analyzers. In this paper, we harness the power of deep convolution neural networks (DCNN) to model sentences and perform sentiment analysis. This approach automates the whole process otherwise done using advance NLP techniques. It is a modular approach analyzing syntactic and context based relation from word level to phrase level to sentence level and then to document level. Such approach not only stands outs in terms of better classification, it also fits the concept of transfer learning. We have achieved an accuracy of 80.69% using this technique and further working on the enhancement and refinement of this approach.