Towards Automatic Detection of Sentiments in Customer Reviews

Abstract

Opinions Play important role in the process of knowledge discovery or information retrieval and can be considered as a sub discipline of Data Mining. A major interest has been received towards the automatic extraction of human opinions from web documents. The sole purpose of Sentiment Analysis is to facilitate online consumers in decision making process of purchasing new products. Opinion Mining deals with searching of sentiments that are expressed by Individuals through on-line reviews,surveys, feedback,personal blogs etc. With the vast increase in the utilization of Internet in today's era a similar increase has been seen in the use of blog's,reviews etc. The person who actually uses these reviews or blog's is mostly a consumer or a manufacturer. As most of the customers of the world are buying & selling product on-line so it becomes company's responsibility to make their product updated. In the current scenario companies are taking product reviews from the customers and on the basis of product reviews they are able to know in which they are lacking or strong this can be accomplished with the help of sentiment analysis. Therefore Our objective of our research is to build a tool which can automatically extract opinion words and find out their polarity by using dictionary,This actually reduces the manual effort of reading these reviews and to evaluate them. The research also illustrates the benefits of using Unstructured text instead of training data which expensive . In this research effort we demonstrate a method which is based on rules where product reviews are extracted from review containing sites and analysis is done, so that a person may know whether a particular product review is positive or negative or neutral. The system will utilize a existing knowledge base for calculate positive and negative scores and on the basis of that decide whether a product is recommended or not. The system will evaluate the utility of Lexical resources over the training data.