AllMusic Mood Subset

The AllMusic Mood Subset (AMS) is a dataset for mood classification from songs. It is created by matching a subset of the Million Song Dataset (MSD), totalling 67k tracks, with expert annotations of 188 different moods collected from AllMusic.

Since the AMS is a subset of the MSD, the audio data is gathered by obtaining the 7-digital 30 second previews associated with all MSD tracks. These are 128kbps mp3 stereo files sampled at 44.1kHz.

Source: Mood Classification Using Listening Data