PerfectMeta: Meta-Data Formating Suggestions

[EliC]

I would like to make this a thread for suggestions / options that would improve PerfectMeta.

-Treat GD3 "[***]" as "(***)" when comparing meta-data. So " [Solo] " from GD3 should be treated the same as " (Solo) " from MusicBrainz or freeDB when comparing to see if there is a match. This is especially important as AMG does not seem to include data from discs in "(***)".

-Option to threat ***/*** as *** / *** when comparing meta-data

-Option to prefer *** / ***

[EliC]

Ex:

5446 Thats My Number / Ball and Chain
vs
5446 Thats My Number/Ball and Chain

There should be an option to ignore spaces on either side of "/" when deciding if meta-data is a match
There should also be an option to decide to prefer xxx / xxx over xxx/xxx



There should also be an option to ignore all spacing when deciding if text matches between sources.

[Porcus]

Just to avoid double posting the entire content from http://forum.dbpoweramp.com/showpost.php?p=77931&postcount=2:
If one (or more) sources are already given priority for a majority of the field, weigh their votes heavier for the rest.

[Porcus]

Ex:

5446 Thats My Number / Ball and Chain
vs
5446 Thats My Number/Ball and Chain

Also, what about & and +, which I presume might be used for two-releases-on-one-CD?

[EliC]

Also, what about & and +, which I presume might be used for two-releases-on-one-CD?

When checking to see if the REST of the meta-data agrees "+", "&", and "and" should probably be treated the same. Then the system should choice between "+", "&", and "and" based on the most common weighted use.

[EliC]

Shouldn't "is" and "it" default to lower case is they are not the first word?

[EliC]

The multi-artist detect algorithm seems so be a bit to aggressive. I have noticed that artists formated as "Some Dude & The Dudets" are detected as multiple artists and remapped to Some Dude; The Dudets. Maybe an exclusion in the rules should be used when &/and is followed by THE ( "& the" or "and the" ).

[EliC]

Is PerfectMeta inserting ", Disc *blooper*" in album titles from multi-disc sets? I am just wondering because if you review the meta-data sources its in ALL of them and they are never that consistent. Is there a way to get rid of ", Disc *blooper*" by either not having PerfectMeta add it (if this is the case) or by having an option to have PerfectMeta delete it if its coming from the data providers.

[Spoon]

You can disable, see meta data options >> Perfectmeta.

[EliC]

You can disable, see meta data options >> Perfectmeta.

found it, thanks

[Vulcan]

You can disable, see meta data options >> Perfectmeta.

Great. I never realized I could disable it.:thumbup:

[EliC]

http://www.dbpoweramp.com/images/perfectmeta-operate.png
.
.
.
Spoon, your example demonstrates what I was saying above. Track 10 AMG and MusicBrainz actually have the same data.
.
.
.

[bhoar]

Spoon, your example demonstrates what I was saying above. Track 10 AMG and MusicBrainz actually have the same data.

Specifically, in post number 2 near top of this thread.

-brendan

PS - why does the pound sign get translated to *blooper* by the board software?

[EliC]

I just found a CD where AMG disagrees with GD and MusicBrainz, who mutually agree -- except for spacing in some (not all) fields. Then since AMG has highest priority, it was chosen for these fields and not for the others. AMG was clearly pointing to a different release too, so I got metadata mixed from two different CDs.

A more "intelligent" (right ...) algorithm could be
-> for the purposes of deciding which ones are wrong, first perform a comparison with spaces and dots and special characters stripped.
-> If two sources agree at that point, then AMG is outvoted
-> If this is the case for sufficiently many fields in one disc, then weight these two sources heavier for the rest.

Example -- disregard freedb here, assume we only use AMG, GD, MusicBrainz:
If GD and MusicBrainz agree that this is Performer 'Foo' Album 'Bar' and AMG says it is Performer 'Someone' Album 'Else', then GD and MusicBrainz have the upper hand. Now if GD writes a song title as 'What If ...' and MusicBrainz as 'What If...', then PerfectMeta would choose AMG's song title. A wiser choice would be to disregard the space and conclude that GD and MusicBrainz agree, and pick one of them.

If GD has 'Foo' and 'Bar!' while MusicBrainz has 'Foo?' and 'Bar' then likewise they should be considered in agreement, I think.


And, finally: Suppose that GD and MusicBrainz are considered to be in agreement on all tracks except the last one, which according to GD is 'Untitled' and MusicBrainz is 'Bonus Track'. Clearly, AMG's information is about a different album than the two others, so AMG should be scored down and either GD or MusicBrainz should be chosen.


Alternatively: If the differences are so significant that it is clear that AMG mentions a different album, then disregard it being outvoted and stick to it -- since AMG metadata can only be inserted upon ripping (while MusicBrainz can later, right?)

re-post from linked thread above.

[EliC]

The multi-artist detect algorithm seems so be a bit to aggressive. I have noticed that artists formated as "Some Dude & The Dudets" are detected as multiple artists and remapped to Some Dude; The Dudets. Maybe an exclusion in the rules should be used when &/and is followed by THE ( "& the" or "and the" ).

Spoon, could you at least separate the options so that the user can only use feat or featuring or ft. The & rule causes more problems then it solves.

[EliC]

With R14 development actively underway I would like to bring this issue up again.

Please posts examples in which perfect meta has failed and ways in which it could be improved to give more accurate meta-data.

[EliC]

Is "featuring" not part of the detect multiple artist algorithm? feat. seems to get picked up, but not featuring.

[Spoon]

Featuring is not.

[EliC]

Featuring is not.

OK, didn't think so. Hopefully R14 will have some updates to perfectmeta, and hopefully this can be added.