Classification and quantification of user's emotion on Malay language in social network sites using Latent Semantic Analysis

Social network sites nowadays serve as important medium of communication and dissemination of information to its users. It is crucial to know users' emotion and perception towards information evolved in social network sites. The motivations for creating tools to detect emotion is increasing due to these factors. Various research conducted recently, focusing on the classification of emotion, that is determining the type of emotion based on certain modality rather than quantifying or recognizing the degree of emotions. The classification is said to determine type of emotion perceived and the quantification is to determine degree of the classified emotion. This paper presents the capability of the Latent Semantic Analysis (LSA) in classifying six basic types of emotion and their three levels of quantification. It is a mathematical technique for extracting and comparing the semantic similarity of meaning between words and passages by analyzing textual information. A prototype is developed to test the accuracy of the prediction based on the proposed framework. Results from the prototype produces 58.3% accuracy for classification and 53.8% accuracy for quantification from randomly selected 173 users' comments. Although, the accuracy is literally low on both classification and quantification, a few drawbacks are identified and highlighted in this paper.