Analysing Amazon product Reviews

This is a project from my Information Retrieval and Data Mining Course while studying MSc Machine Learning at UCL.

How to use

• Place the data file in data/ and specify which category in the main of process_reviews.py.
• Run process_reviews.py to generate the input file for the topic model
• Open topic_modelling.py and update the category accordingly in the main
• Run topic_modelling.py and watch it train. The logger has been set to INFO level so you can see evey set of the training phase
• If you have access to our models, unzip into data. So all models will be in data/models/. No other updates required

Below is an example of how to use topic_model_helpers to analyse the models.

tmh = TopicModelHelpers(['data/models/electronics_20_topics.lda'], model=lda) # load the 20 topics model
tmh.topics # returns a list of topics and their token distribution
tmh.get_reviews_in_topic(19) # show reviews with a proportion of topic 19
tmh.filter_reviews("case", 19) # show reviews that have a proportion of topic 19 and have token case in them

Copyright © 2015 Fayimora. All Rights Reserved.