Feature based opinion mining of online free format customer reviews using frequency distribution and Bayesian statistics

Nowadays, a large number of websites allow users to post reviews about products they bought. There are thousands of reviews of customers related to one product. So it is difficult for the manufacturers and also for the customers to have an idea about the product from these large reviews. In this paper we aim to summarize the customer reviews in factual form as done using two statistical techniques that Bayesian statistics and frequency distribution. The proposed system reads the reviews word by word and finally it finds out the summarized result in terms of frequency and probability of opinions. Bayesian probability is the most current and useful for accurate results and true predictions. Frequency results are in graphical form, the use by a new customer can easily decision to buy the product displayed. Frequency based results can be easily understood by all consumer and the results of Bayesian probability are used to verify the results of frequency. In our experimental results, it proves that proposed statistical techniques are highly effective.