A Linguistic Model for Improving Sentiment Analysis Systems

Abstract

The value of automated sentiment analysis systems is increasing with the vast amount of consumer-generated content, allowing researchers to analyze the information readily available on the World Wide Web. Much research has been done in the field of sentiment analysis, which has improved the accuracy of sentiment analysis systems. But sentiment analysis is a challenging problem, and there are many potential areas for improvement. In this thesis, we analyze two linguistic rules, and propose algorithms for these rules to be applied in sentiment analysis systems. The first rule is regarding how a sentiment analysis system can recognize and apply the semantic orientation of opinion headings in product reviews to features discussed in the review. The second rule we propose allows the sentiment analysis system to recognize informal forms of words used in analyzed documents. Additionally, we analyze the effects of spelling mistakes in text being analyzed by sentiment analysis systems.