Sentiment Analysis using word2vec

3
votes

I am planning to do sentiment analysis on the customer reviews (a review can have multiple sentences) using word2vec. I have certain questions regarding this:

  1. Should I train my word2vec model (in gensim) using just the training data? Should I consider the test data for this too?
  2. How should I represent the review for classification? Will this representation take into consideration the order of the word as this is important in representing a review for sentiment analysis?
1
pythonsentiment-analysisword2vec

1 Answers

3
votes

Basically the answer of your question is already a hot topic of research, here is a research paper that might guide you:

This work is the latest research work I know in this area:

From the paper : Learning Sentiment-Specific Word Embedding for Twitter Sentiment Classification http://ir.hit.edu.cn/~dytang/paper/sswe/acl-slides.pdf

code related to the paper https://github.com/attardi/deepnl/wiki/Sentiment-Specific-Word-Embeddings

Hope this helps!