##SoundCloud Sentiment Research ###Jasdev Singh - Senior Thesis Repository, 2013 - 2014 ###Advisor: Dr. Yanjun Qi

This repo is a work in progress and will serve as central place for code and datasets related to my research in the 2013 - 2014 school year. Four main areas I will be focusing on are:

• Using timed comment sentiment to generate a heatmap associated with a given SoundCloud track
• Clustering timed comments for intelligent previews
• Looking for trends in the BeatPort top charts
• Looking for trends in the DJ Mag Top 100 charts (17 years worth of data)
• Graphing artists' use of other musician's songs in their sets at festivals such as Electric Zoo

All subtasks and goals are open to the public and can be viewed here: https://trello.com/b/Ii6IK0Mi

###Usage To run the SoundCloud sentiment analyzer locally, clone this repo and cd into heatmap-gen/. To install dependencies, run pip install -r requirements.txt. From here, run python sc_runner.py sc_url, to get output on metadata for the track (where sc_url is a valid SoundCloud URL for a given track).

###Structure

• heatmap-gen/comments/ contains the training sets for Bayesian classifier to distinguish sentiment. There are four labelings: negative, neutral, semi-positive, and really-positive. The raw comment corpus is housed in comments.json under this directory
• heatmap-gen/static/datasets contains a few sets of interest. dj-mag-top-100.json and dj-mag-top-100.csv contain the DJ Magazine Top 100 poll results for the past 17 years. Moreover, electric-zoo-2013.json contains set data for the Electric Zoo NY 2013 festival, where each object in the outermost json array contains the artists each DJ 'imports' into their set and rank data.
• aux.py, heatmap_gen.py, top_100.py are all auxiliary files needed to help the web app run and aren't necessary to perform local analysis