Re: [mb-users] Should these AcoustIDs be removed?

On 10/02/2012 04:48 PM, Nicolás Tamargo de Eguren wrote:
>> You can see from the fingerprints that they are all different. It also
>> seems obvious that 083eb0a7-0d6e-44a9-99b0-f2232568ffcc is actually the
>> the Mozart recording.
>>
>>
> I might look silly for asking this, but: how is that obvious? (I have no
> idea how to read that from a big thin vertical strip of black and white and
> wouldn't mind to learn :) )

You don’t actually need to look at the fingerprint to make that
decision, just the number of submissions. On the AcoustID page for the
“correct” acoustID[1], you can see that there are a total of 48
submissions for the that AcoustID and 33 of them have been connected to
a MB recording; all of those 33 are some variation on the same thing ‘Le
nozze di Figaro, K. 492: Act III. No. 20. Duettino ("Sull'aria")’

If you look at the other two acoustIDs, each one has been submitted only
3 or 4 times. Assuming AcoustID is doing its job and assigning different
AcoustIDs to truly different recordings, it seems fair to assume that
with such a degree of difference in submission levels, the one that’s
been submitted the most is assigned correctly.

1. http://acoustid.org/track/083eb0a7-0d6e-44a9-99b0-f2232568ffcc

To get into the less obvious part, looking at the fingerprint images:
Compare the three fingerprints for the single “correct” acoustID. (I
suggest opening each in a different tab so you can switch back and forth
quickly):

http://acoustid.org/fingerprint/12121987
http://acoustid.org/fingerprint/12029311
http://acoustid.org/fingerprint/11752439

 You can see that each fingerprint is nearly indistinguishable, with a
slight time shift or a pixel moved here or there.

Now compare any one of those to a fingerprint from one of the other
*AcoustIDs*:

http://acoustid.org/fingerprint/11752439 (“Mozart”)
http://acoustid.org/fingerprint/10527594 (“High Time”
http://acoustid.org/fingerprint/21764238 (”Celebrate the child”)

They’re not even close to similar; there is no point of reference among
them where you can say “hey, I’m looking at pretty much the same thing.”

*Sometimes* you might find that two different acoustIDs are actually
extremely similar, and it becomes much more of a grey area.

For an example I’ll take “I Got You Babe”[2] by Sonny & Cher. This
recording has two different AcoustIDs:

http://acoustid.org/track/5502d840-500f-48a6-b64c-fb9780bb14a2
http://acoustid.org/track/fd3a066b-5f20-46ec-ae73-3c7a6900802d

I expect that using the most-submitted fingerprint for each is likely to
be the most representative, so we’ll pick those:

http://acoustid.org/fingerprint/11831566
http://acoustid.org/fingerprint/11993528

These fingerprints are pretty similar; You can see a lot of differences,
but the general shape of things is more or less the same — dark parts in
roughly the same areas, same size, same shape.

The degree of similarity is where it gets really difficult: how similar
must they be to be considered “the same” for Musicbrainz purposes?  I
don’t have an answer to that, unfortunately. I would say that these are
not the same, but there’s no definitive line.

2.
http://musicbrainz.org/recording/ffbcb992-bd8d-4410-91af-2a0747955b16/fingerprints

_______________________________________________
MusicBrainz-users mailing list
[email protected]
http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-users

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://www.enigmail.net/

iQEcBAEBAgAGBQJQa28KAAoJEDH4iLg9EM+qwK8H/RL5nsmmUBI9Oi+WMnGXkPIF
CTpiv7QBEmYCEP4J1equJ/shFCIpz2T7xCYVF8x4smoflOO1/5fzEXb+Z5oiIm0t
Xq+kc4Zm+8rhZg9wqT1BDZcyQUdoDwfmi9uFb4S2xHdiWvrFj53zdNvillEHJBRR
JE6kl0YZgSOI3xUnArruv/Ey0istI9sahKPmCsZYU7RXH1kmZ3ZuZITLqX/PT9bE
5WWj2l9jYafgNFZxmDKL4NduKH8PqRzij1oSkq4UADaH9fYhyijfFpWtqSWbTTM9
3Yk6drWXNET+iDlfen2CcXNWA+Dj3UV/bF8PVFoNOyKRp6z1Lha+hlgW4MIk3oo=
=Xpxt
-----END PGP SIGNATURE-----