Emotion Recognition in Assamese Text Using LSTM

Emotion is a crucial part of every human interaction. Emotions can be expressed through written text, speech, body language, or facial expressions. Emotion Recognition is an integral part of natural Language Processing and one of the critical technologies in the human-computer interface in various fields, including Business, Administration, Health care, etc. In recent years, the use of social media and digital marketing has increased, so a massive amount of unstructured data is generated on social media platforms. So detecting emotions from text streams is a challenging task to do. To overcome the different challenges, researchers applied various Machine learning and Deep learning techniques. However, Deep learning methods such as LSTM are highly effective in detecting emotions from the text. In this paper, we use LSTM (Long-Short Term Memory) in Assamese text for better accuracy and prediction. According to recent literature, there is some sentiment analysis work done in the Assamese language. As the Text-based Emotion Recognition research is in the initial stage, our paper aims to detect Emotions from Assamese Text. So, we take two emotion labels Happy and Sad for our research which gives an accuracy of 86% by using LSTM.