API endpoint Audio Features speechiness returning incorrect values

[tazaraz]

The current API endpoint for retrieving audio features, namely '/v1/audio-features', seems to be returning incorrect values. As of this moment, I am unable to find tracks with speech which exceed the 0.33 threshold as per the documentation of the API endpoint field. For example (keep in mind that the API describes everything below 0.33 to probably contain no speech):
 
Gimme Gimme Gimme - ABBA -> 0.0403
Galway Girl - Ed Sheeran -> 0.1
Rap God - Eminem -> 0.314(!)

[volkithedreamer]

Yes, i noticed the same problem. In my playlist, there is no bigger value than 0.37, and there is no correlation with instrumentalness, where there is supposed to be.

[volkithedreamer]

test