Developing a new similarity function for Tribler. Solution used a combination of item-item and user-user similarity, using metadata available in the .torrent files to reduce sparcity in dataset.