Rule-Based Model for Malay Text Sentiment Analysis

Abstract

With the increase number of opinionated content on the web, organizations and people have shown tremendous interest in knowing other’s opinions. This phenomena has attracted both the academic and the business world to pay a close attention towards the development of automated tools which helps in sentiment analysis (SA). While different well-defined approaches have been defined for English SA, the problem remains far from being solved for other languages such as Malay language, despite having more than 215 million Malay native speakers worldwide. To the author’s knowledge, most of researches on Malay language SA rely heavily on the use of bag-of-words model (BOW), which resulted Malay SA to have low accuracy, as BOW model disrupts word order, breaks the syntactic structures and discards some semantic information of the text. In this paper, we propose new feature sets that refine the traditional sentiment feature extraction method and take contextual valence shifters into consideration from a different perspective compared to the earlier research concerning Malay language. The most common valence shifter (VS) are considered in this paper, this includes negation, intensifier, diminisher and contrast. Negation is considered to be the most obvious and common valence shifter of all. A new technique is proposed in this paper to handle complex negation compared to the existing techniques where only simple negations (Bigram) are handled. The proposed system is then compared with existing techniques. The final result showed improvements in Malay SA after considering valence shifter. The discussion and implication of these findings are further elaborated.