• I used BERT (Bidirectional Encoder Representations from Transformers) which is inspired from this paper published by researchers at Google AI Language and is a state of art (SOTA) technique.

• I also used TFIDF and understood why it outperform most the famous architectures like Word2Vec in various cases when used along with Linear Regression as after using the TFIDF it becomes a linear problem and using of algo's like Boosting trees don't work in that case.

• The link to TF-IDF using Linear Regression is

• Visualisation are also made in the same.

• Link to Bert Modelling is this

PS: I also tried using LSTM but it did not give me satisfactory results.