A Semantic Relatedness Approach to Classifying Opinion from Web Reviews

Recent years have marked the beginning and rapid expansion of the social web, where people can freely express their opinion on different "objects", such as products, persons, topics etc. on blogs, forums or e-commerce sites. While the rapid growth of the information volume on the web allowed for better and more informed decisions from users, its expansion led to the need to develop specialized NLP systems that automatically mine the web for opinions (retrieve, extract and classify opinions of a query object). Opinion mining (sentiment analysis) has been proven to be a difficult problem, due to the large semantic variability of free text. In this article, we propose a method to extract, classify and summarize opinions on products from web reviews, based on the prior building of product characteristics taxonomy and on the semantic relatedness given by the Normalized Google Distance and SVM learning. We prove that our approach outperforms the baselines and has a high precision and classification confidence.