A Context-Based Approach for Emotion Recognition From Bangla Text

Abstract

Due to the reliability on various online platforms, numerous texts are spreading all over the world through internet. These texts are employed to enhance user experience, service quality, and other factors. Emotion recognition (ER) is essential in the present day since it is necessary to address user opinion in order to comprehend consumer demand and to properly recognize perceptions. For the purpose of extracting sentiment from various languages, numerous studies have been undertaken. Bangla, a language with limited resources, has recently gained popularity in the field of emotion recognition due to its accessibility on social media. Sentiment analysis (SA) and emotion recognition in Bangla have been applied into a number of works. Multiclass ER has been more widely incorporated in emotion recognition. Though several works have been done on this domain, this language is still resource constrained due to its scarcity of proper tools and techniques. An extensive literature review was used in this study to grasp the state of this field. After that, a compiled framework was developed to serve as advice for those wishing to enter this field. Then, a substantial corpus has been developed to distinguish emotion in a novel approach. The unique aspect of this research is the context-based annotation of the data. After that, emotions are detected known as context-based multilabel emotion recognition. There is no work has been done on this issue. After that, the performance of multilabel ER has been incorporated applying several algorithms. Finally, a web application has been made to acknowledge the performance of the best performer among all the algorithms. Therefore, this research generalizes a roadmap of SA and multiclass ER of existing literature as well as build a comprehensive corpus to detect context-based multilabel ER.