An Empirical Study of the Effectiveness of using Sentiment Analysis Tools for Opinion Mining

Abstract

Sentiment analysis is increasingly used as a tool to gauge people’s opinions on the internet. For example, sentiment analysis has been widely used in assessing people’s opinions on hotels, products (e.g., books and consumer electronics), public policies, and political candidates. However, due to the complexity in automated text analysis, today’s sentiment analysis tools are far from perfect. For example, many of them are good at detecting useful mood signals but inadequate in tracking and inferencing the relationships between different moods and different targets. As a result, if not used carefully, the results from sentiment analysis can be meaningless or even misleading. In this paper, we present an empirical analysis of the effectiveness of using existing sentiment analysis tools in assessing people’s opinions in five different domains. We also proposed several effectiveness indicators that can be computed automatically to help avoid the potential pitfalls in misusing a sentiment analysis tool.