revantkumar / Yelp-Dataset-Challenge-2014

Topic Modeling on Yelp Reviews

User Manual

Configuration used for executing scripts and deploying the application: Python Version - 2.7.6 MongoDB Version - 2.6.5

Installation

• Download and extract the source files from http://a.ashwinikhare.in/yelpolo.tar.gz (Warning : consists of all yelp reviews, File Size > 350 MB)

• The following steps perform various computations to generate Topics & a link of database dump is provided in the end.

  • python Loader.py

    • Depency - pymongo. Run pip install pymongo
    • Populates the reviews for Phoenix from the dataset json where type of business equals to the restaurant

  • python Normalizer.py

    • Dependency - nltk. Run pip install nltk. Also nltk corpus data, So need to import nltk and run nltk.download() in a python shell
    • Makes Reviews more simplified for Online Analysis

  • python OnlineLDA.py

    • Dependency - gensim. Run pip install gensim
    • Gensim python library creates a LDA model for different reviews.

  • python ShowTopics.py

    • Displays the six major topics and the sub-topics with maximum weightages Respectively
    • All 60 topics were categorised so as to highlight the sub-topic they represent
    • The 60 subtopics highlighted in topics.txt

  The required data is now computed and a few auxiliary scripts are present in the .tar.gz which are used to generate other graphs like Business Frequency Topic was plotted by data generated using : getFreqBusinessTopic.py

  The database dump which was computed in our evaluation can be downloaded from http://a.ashwinikhare.in/archive.tar.gz and can be directly imported into the mongodb by adjusting suitable paramters in Constants.py

-Running the Graphical User Interface

- python 'getBusinessTopic.py'
	- Dependency - bottle. Run `pip install bottle`
	- Creates a web server listening at 0.0.0.0:8030 which renders the application
- A live deployed link is present on the server, accessible at http://a.ashwinikhare.in:8030/yelpolo