In other related researches, the term "subject" is not a phrase but is a single word that explains the main content and has been used interchangeably with "theme" or "topic" [13]. An example of subject from a sentence is shown below.

Social tension detection methods using social media textual data have been extensively proposed by the researchers. However, most of the high resourced language which is English. There is limited study on social tension detection methods on low resourced language such as Malay language. In fact, the study of social tension using non-standardized Malay language from social media text such as Twitter is remained unexplored. Textual data on Twitter suffers from inconsistencies and high language ambiguity due to the limited permissible character provided for the users to use. Majority of the existing sentiment analysis systems for social tension detection are based on machine learning approach which depends on the static general-purposes sentiment lexicon and ignores the newly created words on Twitter. As word syntactic on Twitter is dynamically changing according to time and context, machine learning approaches may lead to misclassification of word meaning. This article proposes a lexicon-based sentiment analysis approach for detecting tensions and crime related events indicators on Twitter. The automatic detection of social tension and crime related events of the contents on Twitter helps to discover the indicator of tensions amongst the civilians towards the existing situation and predict the potential crime related events.