musif
musif copied to clipboard
Feature names are ambiguous
To have clearly understandable names, we should define a specific format from which the largest number of information can be gathered. An easy way could be to call them according to a format like:
[module].[FirstOrder|SecondOrder].[Part|Score|Sound].[NameOfPartOrSoundOr-].[Boolean|Int|Float|String].[NameOfFeatureWithoutThePreviousStrings]
e.g.:
SoundVoice_Intervalm3_Per -> interval.2.Sound.Voice.Float.m3_per
where interval
, Sound
, and Voice
are now more well-defined and consequently more meaningful, while 2
and Float
are added
mm yep, sounds like an improvement. Do u think its worth the efforto, or if data its gonna be more usable for data analysis?
Not sure, but this would ultimately help the selection of certain features because it's a clearer format (users can define regex easily) and a better format for pandas API (it works well with multicolumns).
I would leave it here for remembering in future