Sentiment analysis of tamil-english code-switched text on social media using sub-word level lstm

Abstract

Social media are the ultimate platforms to express the opinion and to facilitate the creation and sharing of information, ideas, career interests and other forms of expression via virtual communities and networks. Analysing the sentiment features in these ideas in the public posts of social media users will lead to building more accurate behavioural patterns. Importance of these behavioural patterns with respect to the marketing and business perspective has been focused here. When considering the traditional Facebook marketing platform, efficiency and effectiveness of the marketing are very low since the advertisers do not happen to have a proper understanding of the customers that they should address. Thus, to overcome this issue, a system is proposed to identify the behavioural patterns of Facebook users by analysing their social media contents such as posts, comments, interactions, and also reviews and critics on products to enhance the effectiveness of the Facebook marketing. This system mainly focuses on Facebook users in Sri Lanka. Natural language processing is used to process text-based posts (uploaded and shared) and comments of users in order to build a behavioural profile for the users. This system process text data which is composed by using both English and Tamil languages, in code-switching language pattern.