Sentiment Extraction and Analysis of Product Reviews at Sentence Level

Abstract

These days large number of people are using internet as a medium to communicate and express their views on politics, products, movies and other various domains. Monitoring these views online will be helpful for the companies, individuals or groups, who are interested in knowing what the customers are talking about their products. But manually monitoring is a costly and time-consuming task. Using sentiment analysis one can perform automatic summarization of the customer views based on their opinions. This paper presents work carried out to perform sentiment extraction and analysis of the product reviews at sentence level. Machine learning techniques, namely Multinomial Naïve Bayes (MNB) and Support Vector Machines (SVM)) were used to perform sentiment classification. Experimentation was carried out with two different data corpus and results of the two machine learning techniques were compared. Cross-domain sentiment analysis was also studied and its result is presented. The sentiment analysis system presented here also identifies the product features and extracts the sentiments associated with it. Finally, challenges faced in sentiment analysis and future works are discussed.

Keywords: automatic summarization, sentiment extraction, learning techniques