An experimental approach to compare various deep learning architectures for sentiment analysis

This paper aims to study the efficiency of various seq2seq deep learning architectures for the solution of toxic speech classification and performing efficient sentiment analysis using unilingual publicly available dataset. Numerical examples are presented along with various validation metrics and graphs to indicate the efficiency of the various NLP techniques and confirm the experimental findings of the paper. We also compare and contrast between traditionally used natural language processing models and state of the art model like Bidirectional Encoder Representations from Transformers or BERT.