Last.fm dataset is the official song tag and song similarity dataset of the Million Song Dataset.
Last.fm provides a dataset for music recommendations. For each user in the dataset it contains a list of their top most listened to artists including the number of times those artists were played.
It also includes user applied tags which can be used to build a content vector.
Last.fm dataset provides +940,000 matched tracks of the Million Song Dataset.
I am planning to build a base Recommender System using the number of times certain artists were played as the rating proxy and as a new component, I will try to leverage the user tags to enrich the recommendations from the system by surfacing and matching music gender/category as latent features.