A Comprehensive Roadmap on Bangla Text-based Sentiment Analysis

Abstract

The effortless expansion of Internet access has eventually transformed the dissemination behavior towards E-Mode. Thus the usage of online or, more specifically, ‘Digital’ texts has expanded abruptly. ‘Bangla’, the seventh most spoken language globally, has no different nature. Communication in the Bangla language has also been exposed on the Internet, which describes the feelings of individuals in any specific context. These enormously generated data from diverse sources have drawn the interest of the researchers working in the Natural Language Processing domain. Despite its relatively complicated structure, a lesser amount of annotated data, as well as a limited number of frameworks and approaches, exist. This lacking of resources has kept several stones unturned in this diverse, emotion-rich and widely spoken language. To bridge the lacking and absence of resources, this article aims to provide a generalized deduced working procedure in this domain. To do so, the existing research work in the domain of sentiment analysis using Bangla text has been collected, evaluated and summarized. Also, in this article, the techniques used in pre-processing, feature extraction, and eventually used algorithms have been identified and discussed. Considering these facts, this research work sketches a tentative blueprint of sentiment analysis using Bangla text. Additionally, this article discusses existing regional language corpora such as Tamil, Urdu, and Hindi, as well as English and methodologies used to extract emotional essence from Bangla language comparing other languages. That will assist in determining the probable chosen path of exploring Bangla in a more deeper aspect. Moreover, this work has deduced and presented a generalized framework that will direct aspiring researchers to decide the pathway of choosing data vis-à-vis methodologies based on their interests.