Awesome

Sentiment Analysis for Amazon Product Reviews

Task

• Dataset : 400 thousand reviews of unlocked mobile phones sold on Amazon.com
• Problem : Sentiment analysis for Amazon product reviews
• Use Natural Language Procesisng techniques, Bag of Words model, Word2Vec model and Long Short Term Memory (LSTM) neural network to conduct sentiment analysis for Amazon product reviews.
• Accuracy score of 94.4% by Word2Vec embedding with LSTM.

What is in this repo

amazon-sentiment-analysis.ipynb

• Data visualization of Amazon product reviews
• Preprocess raw reviews to cleaned reviews
• Use different word embedding models, such as count vectorizer, tf-idf transformation and Word2Vec model, to transform text reviews into numerical representations
• Fit numerical representations of text reviews to LSTM (a type of recurrent neural network)

For more detailed information, see Report.