Sentiment analysis in medication related texts

This study deals with the sentiment analysis of medication related texts improvement by using classical and state-of-the-art approaches like long-short-term memory (LSTM) and bidirectional encoder representation from transformers (BERT) model. The models evaluation demonstrates that transformer model outperformed other models like Nave Bayes (NB), logistic regression (LR) and support vector machine (SVM). Despite the superiority of state-of-the-art models, we show that the use of classical approaches remains a possible choice to avoid the complexity problem of the state-of-the-art approaches. The texts used in our experiment come from the Social Media Mining for Health (SMM4H) corpus and the coronavirus tweets NLP-text classification corpus.