Multi-label Emotions Bangla Text Classification

Abstract

Emotion recognition in text is a challenging task, especially in languages with complex syntax and semantics such as the Bengali language. This paper presents a novel approach for multi-label emotions in Bangla text classification, aimed at accurately identifying and categorizing emotional content within Bengali texts especially encompassing emotion of happiness, sadness, anger, fear, and surprise, among others. The proposed method leverages advanced natural language processing (NLP) techniques, including word embeddings, deep learning models, and multi-label classification algorithms. The first step involves preprocessing the Bangla text data, including tokenization, stemming, and removal of stop words to enhance the quality of feature extraction. We then utilize pretrained word embeddings to represent the semantic meaning of Bengali words effectively. Subsequently, a deep learning architecture, such as a Bi-LSTM(BiLSTM) network will employed to capture the contextual information and hierarchical relationships within the text. To address the multi-label nature of emotions in Bangla text, we will try to employ techniques such as binary relevance, classifier chains, and label powersets to enable the classification of multiple emotions simultaneously. We achieved the most optimal result at a training accuracy of 89.76%, a training loss of 29.88%, a validation loss of 35.22%, and a validation accuracy of 22.84%. We will also evaluate the performance of our proposed approach using standard evaluation metrics precision, recall, and F1-score. Experimental results will demonstrate the effectiveness of the proposed method in accurately classifying multi-label emotions in Bangla text. The model achieves competitive performance compared to existing approaches, indicating its potential for practical applications in sentiment analysis, opinion mining, and affective computing in Bengali-language domains. The findings of this research can contribute to the advancement of emotion recognition technologies in linguistically diverse contexts, paving the way for a more nuanced understanding and interpretation of textual emotions in the Bengali language