kalli / essentialmixstats Goto Github PK

I noticed some inconsistencies in the original data, which may affect statistics. For example, Y-Traxx and Y Traxx are considered different artists, and so in the Most Played Tracks section, there are 2 separate totals for the same track "Mystery Land" (7 and 6 respectively) which should actually be combined.

To try and normalize these, I created a simple script which attempts to find similar artist names by removing punctuation, whitespace, and diacritics, and then looking for other artist names with a Levenshtein distance <= 2. output & source

The output is very raw... and it needs to be curated by hand; there are many false positives hiding in the data.

I hope you can adapt my sample code and use the output to pre-process the data so that the totals are more accurate.

Thanks for doing this work, it's really interesting!