Sentiment Analysis of User Posts

Project Summary:

  • Implemented two methodologies and compared them * Firstly, using a lexicon-based approach * Secondly using machine learning techniques like SVM, Logistic Regression, artificial neural networks.
  • SVM performed the best of all, due to the constrained size of training dataset used.

Recent Developments:

  • Recent Development: Performed sentiment analysis of yelp reviews using LSTM, Bi-directiional LSTMs, Convolutional Neural Networks.
  • Simple LSTM model produced the best accuracy. Suprisingly, the Bidirectional LSTM overfits on the data does not produce better results compared to the LSTM Network.

FrameWorks : Keras, Tensorflow, scikit-learn, NLTK