A New Term Weighting Method by Introducing Class Information for Sentiment Classification of Textual Data
Long‐Sheng Chen, Chia‐Wei Chang · 2011
Abstract—With the popularity of text based communication tools such as blogs, Plurk, Twitter, and so on, customers can easily express their opinions, reviews or comments about purchased products/services. These personal opinions, especially negative comments, might have a significant influence on other consumers ’ purchasing decisions. Therefore, how to detect users ’ sentiment from textual data to assist companies to carefully respond to customers ’ comments has become a crucial task. Recently, machine learning methods have been considered as one of solutions in sentiment classification. When applying machine learning approaches to classify sentiment, Term Frequency (TF), Term Presence (TP) and Term Frequency-Inverse Document Frequency (TF-IDF) usually have been employed to describe collected textual data. However, these traditional term weighting methods cannot have positive influence on improving classification performance. Therefore, this work proposes a new term weighting method called Categorical Difference Weights (CDW) by introducing class information. Besides, CDW will be integrated into Support Vector Machines (SVM). Finally, an actual case will be provided to illustrate the effectiveness of our proposed method. Compared with traditional term weighting methods, TF and TF-IDF, experimental results indicated that the proposed CDW method indeed can improve the classification performance.