Vietnamese Online Hotel Reviews Classification Bases on Term Features Selection

Abstract

We propose an improved technique to analyze sentiment for Viet-namese texts based on term feature selection approach. The sentiment analysis task is to classify a sentence into one of the following predefined categories: positive, negative, and neutral. In order to analyze the sentiment, we compare three different text categorization algorithms including Decision Tree, Naive Bayes (NB) and Support Vector Machines (SVM). Furthermore, we enhance the efficiency of the text categorization by applying feature selection technique, χ²(CHI). The evaluation was conducted on 1,650 hotel reviews written in Vi-etnamese languages. The experimental results showed that applying term feature selection could significantly improve the performance of the sentiment analysis.