Valence as a measure of happiness

[DonSolare]

In Spotify's API is something called Valence, that describes the musical positiveness conveyed by a track. Tracks with high valence sound more positive (happy, cheerful, euphoric), while tracks with low valence sound more negative (sad, depressed, angry).

But how can Spotify's API measure a "happy feeling" with numbers? can you give me some additional information about the algorithm involved? Thanks!

[erichie]

This is an attribute that we compute based on a wide variety of inputs. Generally speaking, we use a set of agreed upon ideas of what happy or positive music sounds like.
This was originally developed at The Echo Nest.
For more, read:
https://beta.developer.spotify.com/documentation/web-api/reference/tracks/get-several-audio-features...
https://web.archive.org/web/20170422195736/http://blog.echonest.com/post/66097438564/plotting-musics...

[caitlina88]

I was wondering if someone could elaborate more on this topic? I visited both links and none of them say specifically how valence is measured, other than by "advanced-computer-learning." How is Spotify defining valence? Minor/major key? Lyrical content? Any information you can provide would be greatly appreciated. 

[alejandroheca]

Hi! Did you ever get an answer to this issue? I am working on a research project and we are also interested in using the valence.

[caitlina88]

Hey there. No, I never got a response, sadly. If you learn anything about this measure, I’d love to hear about it!

[serkansulun3]

My guess is that they run a pretrained model on the audio samples. There are datasets that include the raw audio and human labeled valence/arousal values, so it is possible to train a model, most likely a (deep) neural network.