Class Projects - Spring 2013


MusixRx makes music recommendations for users based on their Twitter activity.

MusicRx project from the Collaborative Innovations class Spring 2013.
What it does

Content recommendation engines are a huge business and a huge challenge to create. Most of recommenders function by looking at a consumer’s purchasing history, rankings of other media, or stated preferences. MusicRx uses Twitter as a data source and makes recommendations based a user’s Twitter history by analyzing the words, hashtags and @-mentions she uses most. MusicRx is similar to our BookRx project. 

How it works

Using MongoDB, MusicRx keeps a database of Twitter users with known musical preferences that includes the words, hashtags, and websites mentioned most frequently in their recent tweets.

When a MusicRx user searches their own Twitter handle, the system retrieves their most recent tweets and runs the nearest neighbor classifier to compare the words and other elements in their tweet history against the matrix of known users. Twitter users with similar tweet histories are presumed to have similar tastes in music, which enables MusicRx to return recommendations.

Next Steps
  • Improve speed by implementing better code.
  • Improve accuracy of recommendations by expanding preference database to identify musicians that people follow using automated Twitter data from other music services like Pandora.

Screenshot of the user interface.

Diagram showing the data flow and analysis MusicRx uses.


Notes from the team who built MusicRx

Initial Concept: Larry Birnbaum and Shawn O’Banion

Student Team: Alex Wendland (journalism), Vanessa Fang (computer science), Tommy Yu (computer science)

Faculty Guidance: Larry Birnbaum, Rich Gordon and Kris Hammond (with assistance from Shawn O’Banion)