Is there any explanation for the meaning of the lastfm emotion tags?

[Aviram-A]

I noticed it's all values between 0-100, but how were they calculated?

[TuringTrain]

Hi,

at the end of page 5 of our paper (https://arxiv.org/pdf/1912.02477.pdf) this is explained and the relevant papers are cited:

Based on social tags from lastfm, a list of social tags that are related to emotion are identified, then projected into the valence-arousal space using an emotion lexicon.

For example, a song is tagged as "rock, upbeat, happy, rainfall". Then the emotion-related subset of tags {upbeat, happy} is identified and the song's "emotion" is modeled as (valence, arousal) based on the valence and arousal of those tags in the dictionary. In this case, high valence and arousal, based on the (val, aro) of {upbeat, happy}.

[Aviram-A]

Thanks for replying! I went over your answer and the article sections you mentioned, but I still don't understand how the values in the LastFM emotion tags (from the readme file) were obtained.

The way I see it, every song is initially tagged with one or several emotion tags. How were these tags transformed to values between 1-100 that are available in the dataset? What do these values mean?

Thanks again!