Thumb Up or Down? A Text-Mining Approach of Understanding Consumers through Reviews

Word of mouth has long been recognized to be an influential variable in marketing. With the growth of Internet applications, traditional word of mouth has evolved into the online form in a variety of Web-based outlets where individuals spread their perceptions via the written word. These expressions are often in the form of online reviews or assessments of products and services. In this article, we attempt to use features to represent reviews, which contain the sentiments of the consumers, and to predict the overall attitudes of online reviews of the consumers. Further, we want to look at which words are indicative/decision driven of a positive/negative attitude of the consumers, especially we want to identify a set of features which will result in a desired class-positive attitude in our case. Data was collected from a well-known web site using a WebCrawler type technique and we applied text-mining approach for the analysis. The overall results compare favorably with those from standard numeric based quantitative prediction methods. In addition, the text-mining methodology and inverse classification help us identify the key features that are related to positive/negative overall attitudes of online users. Identification of key features will be of considerable help to marketers in designing their keyword choices for more effective application of search engine marketing strategies while identification of the negative associated key words will lead to discovery of problematic areas.