Uploaded image for project: 'Zapped: AcousticBrainz'
  1. Zapped: AcousticBrainz
  2. AB-470

Allow to either collapse or otherwise hide same MBID in similarity lists

    • Icon: Improvement Improvement
    • Resolution: Unresolved
    • Icon: Normal Normal
    • None
    • None
    • None
    • None

      https://acousticbrainz.org/similarity/d17e53d3-6995-43a6-a2fd-f526a4b691f1/moods?n=0 has "Spyro Gyra - Escape Hatch" listed multiple times for various different submissions of the same. While the individual submissions’ similarity can absolutely be of interest (e.g., to check for wrongly merged submissions or other more nerdy things), it is likely not the most interesting if used for music discovery/recommendation purposes.

          [AB-470] Allow to either collapse or otherwise hide same MBID in similarity lists

          Freso added a comment -

          I don’t really know, tbh. I think for music discovery or recommendation, it would make sense to only show it once (but probably weighted so if 1 submission would rank #2, the next 4 submission would rank #5, and then the final 2 would rank #7, probably put the whole thing in #5?). For the more nerdy purposes of checking for MBID (mis)identification purposes, it would probably make sense to group them and list them separately? Or maybe not if you can still "uncollapse" the hidden ones.

          Freso added a comment - I don’t really know, tbh. I think for music discovery or recommendation, it would make sense to only show it once (but probably weighted so if 1 submission would rank #2, the next 4 submission would rank #5, and then the final 2 would rank #7, probably put the whole thing in #5?). For the more nerdy purposes of checking for MBID (mis)identification purposes, it would probably make sense to group them and list them separately? Or maybe not if you can still "uncollapse" the hidden ones.

          I had another discussion with some people about this, we decided that we'd remove duplicates (with the same MBID) from this list, and show a "And also 7 other duplicate submissions (show them)" link which allows people to see them.

          One thing that we noticed in the API results is that there are two different types of duplicates - ones where the distance is exactly the same, and ones where it's different (with possibly other mbids in the middle). Do you think that we should skip all mbids that are the same, showing just the closest one, or only collapse non-interrupted sequences of the same mbid?

          Alastair Porter added a comment - I had another discussion with some people about this, we decided that we'd remove duplicates (with the same MBID) from this list, and show a "And also 7 other duplicate submissions (show them)" link which allows people to see them. One thing that we noticed in the API results is that there are two different types of duplicates - ones where the distance is exactly the same, and ones where it's different (with possibly other mbids in the middle). Do you think that we should skip all mbids that are the same, showing just the closest one, or only collapse non-interrupted sequences of the same mbid?

            Unassigned Unassigned
            freso Freso
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:

                Version Package