A Study of Various Word Embeddings in Deep Learning

With the number of users and reviews rising on the internet, and the liberty was given to post anything on it, it is still a major issue to identify the sentiment of the reviews. The traditional machine learning models are taking a lot of computational time to get desired results. The advancement in the field of deep learning has left people wondering whether it would provide better results for such applications. Word Embeddings play a huge role in the classification of text, as they represent similar meaning documents with equivalent representation. The result of the deep learning model trained on a particular set of data depends upon the sort of word embedding used. In this work, we intend to present a study of various embeddings like GloVe, BERT, Word2vec, and Fast-text using a proposed model. The research was conducted on the IMDB movie review dataset and the result displayed that GloVe embeddings produced the best result with an accuracy of 87.936 % on our proposed model.