Emotion Detection from Text: Classification and Prediction of Moods in Real-Time Streaming Text

Emotion detection is a method which can be used to determine publics' attitudes, feelings, and feelings towards a specific target, such as persons, groups, organizations, various services, and products. Sentiment analysis is a superset of emotion detection since it infers the specific emotion rather than just declaring if something is good, bad, or neutral. Recent studies have concentrated on the use of verbal and facial clues to identify emotional states. Since written language lacks non-verbal indicators like voice tone, face expression, vocal pitch, etc., it can be challenging to discern emotions. The extraction of emotions from text has been proposed using a variety of natural language processing (NLP) techniques, such as the keyword-based approach, the lexicon-based approach, as well as deep learning approach. The disadvantages of keyword and lexicon-based approaches are numerous despite their focus on semantic links. In this research we suggest a BERT-based deep learning system (Fake BERT) by combining the BERT with numerous parallel blocks of a single-layer deep Convolutional Neural Network (CNN). GRU and Bi-GRU comparisons from past works were used as deep learning approaches. The proposed BERT-1D CNN generated the best results with an F1 score of 83.5, followed by Bi-GRU and GRU.