Evaluation of Product Reviews by Natural Language Processing Using Machine Learning

Abstract

The review function of the internet shopping is the function which the purchaser can write impression and evaluation of the goods. It is possible that the purchase applicant reads the product review, and that it obtains information of the product beforehand. However, some product reviews are not helpful because all buyers do not write reviews carefully. In this study, we analyze and evaluate product reviews by natural language processing using machine learning, and construct a system that rearranges products in order of reference. Learning is carried out using logistic regression so that the words which appears in the content of the review many times is made to be the learning feature, and the appropriate weight in which the learning feature affects the content of the product review is calculated. For the evaluation of the system, the QE method which is the evaluation method using the quick sort is proposed. By using the QE method, it was possible to evaluate whether the rearrangement was right. As a result of the experiment, it was confirmed that the proposed system could rearrange the product reviews with high accuracy.