Recommender Systems and Personalised Multimedia
MyMedia was an EU-funded collaborative research project which developed state-of-the art recommender algorithms.
What we've done
The MyMedia project developed recommender algorithms for multimedia content and evaluated them in a series of field trials which addressed different application areas.
The BBC field trial addressed online catch-up services for TV and radio.
Gender classification, radio art, surprisal, and creativity
More project info
Why it matters
The large number of available channels and emergence of on-demand systems have created a "Crisis of Choice" for users. Successful broadcasters will be those who make it easy for users to find the content which is most likely to appeal to them.
- To help people find interesting TV and radio programmes
- to explore the potential value of user behaviour data
How it works
The MyMedia algorithms use collaborative filtering, where personalised recommendations are obtained by analysing the collective behaviour of users.
Collaborative filtering provides good results but is difficult to implement in a broadcasting environment, where there is a constant flow of new content.
MyMedia addressed this issue by providing algorithms which support incremental updates.
The MyMedia project improved our understanding of collaborative filtering systems and we gained valuable experience of how these can be applied within a broadcast environment. The algorithms developed by the project won several competitive events:
Context-aware Movie Recommendation Challenge (CAMRa 2010)
These algorithms are now available in the MyMediaLite Recommender System Library which started within the project and continues to be developed.