Quora Insincere Questions Classification

Tested Approaches:

I. Probabilistic Modelling

1. TF-IDF based Vectorization w/o Lemmatization + Logistic Regression
2. TF-IDF based Vectorization w/o Lemmatization + Multinomial Naive Bayes Classification

II. Deep Learning

1. Selected Hyperparameter values for max_features (Number of words learned by Keras Tokenizer) & maxlen (Maximum permissible Length of Sequence).
2. Unpacked GLoVe (Global Vectors for Word Representation) Embeddings for Keras API's Embedding Layer
3. Trained a Recurrent Neural Network using GLoVe Embeddings and obtained an accuracy of 95.62% during Cross Validation.