Comprehensive Analysis of Drug Reaction Opinions Through Deep Learning

Now days social media is growing exponentially and people post their opinions upon different platforms. We are concerned about patient's opinions related to the medicines or drugs. Drug reaction opinion is one of the most important dimension to evaluate the performance of the healthcare system. Sentiment analysis is a field of NLP that provide the meaningful information about sentiment, emotion or attitudes from reviews. In this article we have used Deep learning based opinion mining to extract the most relevant features. We embedded our inputs by two types of embedding technique - Glove and Word2Vec. These embedding passed in deep learning methods like CNN, GRU, LSTM, RNN, and CNN+LSTM. These models have good performance. We first proposed SWIN transformer model that is used in image processing. This model provides state-of-the-art performance, and then we compared it with other transformers like BERT, DistillBERT, RoBERTa. We also compare SWIN performance from the previous research which is still lacking in the performance in terms of accuracy.