This is to create models for Tweet sentiment, Yelp, Amazon echo review classification with NLP techniques.
Multi-nominal naive bayes and Logistic regression models are used.
Data visualization, feature extraction via CountVectorizer are used, and TfidfTransformer is used to further extract features.
Tweet sentiment model(s) are created using Deep CNN in combination with Bert tokenizer and Bert embedder, respectively for sentiment analysis. Embedder based model shows good generalization than tokenzier model.
Production and deployment level codes are to be opened soon. Stay tuned.