Emotion or sentiment recognition using text analysis

Abstract

Natural language processing has become popular in the recent years since the evolution of artificial intelligence. Emotional and sentiment analysis, a subfield of natural language processing is used to detect and predict emotion of a text phrase. This report explores the usage of multiple deep learning models, such as Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), Bidirectional Gated Recurrent Unit (BiGRU), CNN-BiLSTM and CNNBiGRU and an ensemble of the best 3 models. This experiment will enhance their accuracy for text prediction based on Ekman’s six basic emotions Sadness, Joy, Love, Anger, Fear, Surprise and comparing their performance to the base configuration of the transformer models such as distilBERT, BERT and roBERTa. Different parameters and deep learning model configurations are experimented on to test the best accuracy, precision score, f1 score and recall of each model and compare their computational cost.