Although research on emotion classification has significantly progressed in highresource languages, it is still infancy for resource-constrained languages like Bengali. However, unavailability of necessary language processing tools and deficiency of benchmark corpora makes the emotion classification task in Bengali more challenging and complicated. This work proposes a transformer-based technique to classify the Bengali text into one of the six basic emotions: anger, fear, disgust, sadness, joy, and surprise. A Bengali emotion corpus consists of 6243 texts is developed for the classification task. Experimentation carried out using various machine learning (LR, RF, MNB, SVM), deep neural networks (CNN, BiLSTM, CNN+BiLSTM) and transformer (Bangla-BERT, m-BERT, XLM-R) based approaches. Experimental outcomes indicate that XLM-R outdoes all other techniques by achieving the highest weighted f 1 -score of 69.73% on the test data. The dataset is publicly available at https://github.com/omar-sharif03/ NAACL-SRW-2021.
Emotion classification in text has growing interest among NLP experts due to the enormous availability of people’s emotions and its emergence on various Web 2.0 applications/services. Emotion classification in the Bengali texts is also gradually being considered as an important task for sports, e-commerce, entertainments, and security applications. However, It is a very critical task to develop an automatic emotion classification system for low-resource languages such as, Bengali. Scarcity of resources and deficiency of benchmark corpora make the task more complicated. Thus, the development of a benchmark corpus is the prerequisite to develop an emotion classifier for Bengali texts. This paper describes the development of an emotional corpus (hereafter called ‘BEmoC’) for classifying six emotions in Bengali texts. The corpus development process consists of four key steps: data crawling, pre-processing, labelling, and verification. A total of 7000 texts are labelled into six basic emotion categories such as anger, fear, surprise, sadness, joy, and disgust, respectively. Dataset evaluation with 0.969 Cohen’s
κ
score indicates the close agreement between the corpus annotators and the expert. The analysis of evaluation also represents that the distribution of emotion words obeys Zipf’s law. Moreover, the results of BEmoC analysis shown in terms of coding reliability, emotion density, and most frequent emotion words, respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.