Sentiment-specific word embedding for Indonesian sentiment analysis

Word embedding has been proven to improve performance of sentiment classification systems. A new proposed word embedding called sentiment-specific word embedding exploits sentiment information, in order to reach better performance of sentiment analysis systems using word embedding. In this paper, we build Indonesian sentiment-specific word embedding and apply it for sentiment analysis. We compare performance of sentiment classification based on sentiment-specific word embedding with other popular feature representations such as bag of words, Term Frequency-Inverse Document Frequency (TF-IDF), and also generic word embedding. Although our sentiment-specific word embedding models achieve a higher score than the Word2Vec embedding, the lexical representations, bag of words and TF-IDF still gained better performance.