Sarcasm Detection followed by Sentiment Analysis for Bengali Language: Neural Network & Supervised Approach

As www data grows, so do opinions, views, visitors, news, and comments. Using opinions, perspectives, and remarks, Natural Language Processing (NLP) professionals may classify emotions. Classifying and evaluating Bengali text emotions is becoming significant in e-commerce, journalism, movies, OTT, and security applications. The lack of Bengali corpus makes developing a Sentiment Analysis system difficult. Sarcasm is another popular social media trend. Positive words are often used to indicate hatred. Thus, it’s hard to tell what these sentences mean. This study presents a method for identifying and analysing sarcasm. GloVe is used to represent words while LSTM is trained and tested on the represented characteristics. Experiments show 91.94% accuracy. Predicted sarcastic sentences are labelled as negative and added to Sentiment Analysis corpora (SA). Logistic Regression (LR), K-Nearest Neighbor (K-NN), Linear Support Vector Machine (SVM), and Random Forest (RF) are used to feature matrices for sentiment analysis. For Unigram, Bi-gram, and Tri-gram models, Linear SVM has the highest precision (92.5%), whereas LR model approaches greater accuracy (72.04%) and F1-score (68.15%).