/sentiment-anlysis

Using neural networks with pre-trained word embeddings such as GloVe and word2vec to predict sentiment for given text.

Primary LanguageJupyter Notebook

Description

This project aims to study the sentiment analysis using Deep Neural Networks (CNNs and RNNs). During process, we will:

  • Process text data.
  • Explaine the importance of word embeddings and the powerful idea of embedding stuff.
  • Experiment public available word embeddings (GloVe and Word2Vec).
  • Use Convolutional Neural Networks to train a sentiment classification model which is better than state-of-the-art resutls of other methods such as Latent Semantic Analysis.
  • Experiment Recurrent Neural Networks, more specifically, the long short term memory networks (LSTM).

A very detailed explaination along with accompanying readable code is available in the notebook.

Tools

  1. Python 2.7 and dependencies: numpy, pandas, bcolz, pickle, json, nltk, Theano or Tensorflow, Keras, re

programmed and documented in Jupyter notebook.

  1. Train on a GPU Testla K80 4x Server or, equivalently, Amazon Web Service EC2 instance (p2.xlarge).