A textual polarity analysis based on reviewer identity disclosure and product sales

Analyzing the emotional polarity of unstructured text is an important research topic in sentiment analysis and has attracted much attention in the past few years. In our work, in order to analyze the emotional polarity of text, we consider using economic techniques instead of manual annotation and linguistic resources. The fact is relied on that textual polarity will affect the subsequent consumer behavior which would affect the product sales and consumer identity disclosure in comment. This influence can be observed by using some easy-to-measure economic variables such as product price or product sales. Reversing the above logic, we can infer the textual polarity the by tracing reviewer identity disclosure and product sales. We will propose a regression model to analyze the textual polarity effectively without the need for the manual labeling. The discussion is made by presenting results on the reputation system of Amazon.com. The results show that we can infer the textual polarity by measuring reviewer identity disclosure and product sales.