Jaikoz Automatic Music Tagger New version - 8.3.5

Just a small release (8.3.5) to fix some recently found bugs in Jaikoz with song identification

• [JAIKOZ-1065] - Query by tracks needs an upper limit on how many tracks can be queries in one query
• [JAIKOZ-1067] - Invalid attempt to get recordingId rather than trackId for Discogs check
• [JAIKOZ-1068] - IsGoodArtist check not working well when comparing colloboration to one artist

Awesome song identification with SongKong 3.17

Over the Summer months we released new versions of Jaikoz, culminating with the ultimate song identification

Now we have put all those improvements into the latest version of SongKong - and we release it today, we have even squeezed in a few more improvements that are not yet in Jaikoz.

We have also had the Spanish and German in program translations reviewed and corrected by professional translaters, if you find any issues  just let us know.

Now we are really keen to get support for new audio formats and new features into both SongKong and Jaikoz over the coming months.

The full list of fixes can be found here

Jaikoz 8.3.4 Released

We have just released Jaikoz 8.3.4

This release fixes an issue with Discogs matching plus two others:

[JAIKOZ-1009] - Export to xls exportsComent language for Xls causing problem for Import
[JAIKOZ-1059] - OSX:Error opening files in Finder with right-click unless Jaikoz already running
[JAIKOZ-1060] - Incorrect check being done for Discogs matches 

We've done it:Ultimate Song Identification

A couple of months ago long time Jaikoz user Alfg reported song identification was not working as well as in the previous major version. We did some investigations and quickly found a couple of issues that increased the number of songs identified.

But there was still some songs that were in the MusicBrainz database and a few songs that were being misidentified. It got me thinking is it possible correctly identify every single song if its in the database, and not misidentify any.

It is a difficult tightrope to walk trying to identify more songs where there is not a clear answer, ensuring we do not misidentify songs, because this is worse than not identifying the song at all.

It has taken a few iterations but now with Jaikoz 8.3.3 for these particular tests we now have 100% correct matches and no mismatches for a series of essentially random folders.


 

I realize not all our customers will have such success but I think most of you will be very happy with the improvements we have made, now I need to incorporate some of these improvements into the next version of SongKong.

The improvements we made in this release are as follows:

• [JAIKOZ-1043] - isGoodArtistAndTitleMatch failing artist which just has additional 'the' in name
• [JAIKOZ-1044] - Single song match doesnt do release match if metadata has song title
• [JAIKOZ-1047] - ManualCorrect isAcoustid checkbox is never set
• [JAIKOZ-1048] - Discogs matching should enable track duration check
• [JAIKOZ-1049] - Only allow recordings with same name AND SIMILAR DURATION to one with most sources
• [JAIKOZ-1053] - Matching to multidisc release can fail to match disc 2 because false match on disc 1
• [JAIKOZ-1054] - Simplify album names when trying to find releases by metadata
• [JAIKOZ-1057] - Try harder to find artists when grouping songs that have no artist

More Jaikoz matching improvements with Jaikoz 8.3.2

Jaikoz 8.3.2 Released 

 Some more tweaking to the song identification algorithm plus fixing some recently discovered bugs and we have another new release:

Bugs

• [JAIKOZ-0999] - Error with highlighter when drop folder into Jaikoz
• [JAIKOZ-1037] - NullPointer with HighlightDuplicateMusicBrainzUniqueIds
• [JAIKOZ-1039] - java.util.NoSuchElementException in ArtistTitle check
• [JAIKOZ-1040] - OSX:Error copying album artwork directly from Google Chrome

Improvements

• [JAIKOZ-1023] - What do do if tie and onlyUseRecordingWithMostSources enabled
• [JAIKOZ-1029] - Ignore track duration check if track was already matched by Acoustid
• [JAIKOZ-1034] - Double check songs by metadata only matches checking artist and title
• [JAIKOZ-1035] - Increase artist/title check for single song metadata match

Not so perfect

Okay, a little regression slipped into the 8.3.0 release but now fixed and deployed in Jaikoz 8.3.1

• [JAIKOZ-1033] - Artist/title double check for single song metadata matching is incorrect

The Perfect Song Matching Algorithm ?

The quest for the perfect matching algorithm continues, but we are certainly getting closer. With newly released Jaikoz 8.3.0 we have made many improvements to the matching process.

We now do more to groups songs correctly and find possible potential releases, such as searching for releases by track titles if no matches found by release title.

And we are now stricter about accepting a match, we have added some additional checks to ensure it really is a good match.

So the match rate is now higher, and at the same time incorrect matches are kept to a minimum.

• [JAIKOZ-773] - Allowing Acoustid matches when metadata exists but is bad
• [JAIKOZ-1005] - If artist name contains feat and get no matches strip out and try again
• [JAIKOZ-1007] - When single song matching against Acoustid ensure we are selecting the correct song
• [JAIKOZ-1013] - Dots not removed from filepath when after path separator
• [JAIKOZ-1014] - When disc has more than two discs may not match because can match songs to wrong disc
• [JAIKOZ-1018] - Ignore join phrased when matching artists
• [JAIKOZ-1020] - Consider artist name when doing track scoring
• [JAIKOZ-1022] - Always only use recording with most sources when single song matching
• [JAIKOZ-1024] - Extend Acoustid title check to an artist check
• [JAIKOZ-1026] - Additional processing when songs organized one folder per artist
• [JAIKOZ-1027] - Search by metadata should search by artist and tracks if no match on release

Better song identification with Jaikoz 8.2.5

In this release we have spent some time working out why some songs by Jaikoz do not get matched even though they exist in the Albunack database and we found a few bugs and improvements we could make.

• [JAIKOZ-1003] -         Single song matching only matching by acoustids
• [JAIKOZ-1004] -         Single song matching by Acoustid should only use recording with most sources
• [JAIKOZ-1005] -         If artist name contains feat and get no matches strip out and try again
• [JAIKOZ-1010] -         Unable to create fingerprints on Windows if filepath greater than 260 characters

So now you should get an even better matching rate with Jaikoz, already we believe have the most sophisticated and accurate matching algorithm out there, if you know any different let us know.

       

Better song matching for songs of random folders with Jaikoz 8.2.4 Music Tagger

Jaikoz 8.2.4 released June 29th 2015

A new release of  Jaikoz with an important fix

Jaikoz relaxes its matching criteria if it detects a folder that  contains songs not limited to one album and artist. Whilst you may have some of your songs organized one folder per album its not uncommon to have  folders containing songs from all kind of sources in no discernible order, fixing these in Jaikoz is why you bought it in the first place, right ?

If Jaikoz identifies such a folder it relaxes its matching criteria so even if Only Match all songs to one album if all songs in grouping were matched is enabled it is ignored allowing these songs to be matched without breaking up folders that have already been organized

But there was a bug preventing it properly matching songs in such a folder even if the option was disabled. Now with this fix it does its matching correctly for these types of folders, even if the option is enabled. So you should now get much better results when trying to match such folders.

Reports, Error handling and iTunes improvements with SongKong 3.16

We have a new release of SongKong, version 3.16 !
  
This is a small release with some bug fixes and minor improvements

iTunes integration

This release fixes two issues with SongKongs iTunes integration:


SONGKONG-863:OSX:Error when unable to find track in iTunes

SONGKONG-864:OSX:ITunes integration can cause Songs Done to be under-reported

Error Handling

Usually SongKong can be run on any number of files without memory issues, but badly behaved images can cause a memory issue. SongKong was not recovering from this error and if the memory issue did reoccur when loading songs SongKong was not reporting the error correctly, this is now resolved. 

SONGKONG-868:If memory error occurs on file load with cover image doesnt recover properly

Report Improvements

We have had added various report improvements

SONGKONG-494:Add an option to not show reports

We have added a new option for reports. This option can be used to prevent the report being opened automatically, the report is still created and can be opened immediately after processing from the summary panel or from the Reports menu.

SONGKONG-869:Progress Dialog should close automatically once finished processing

Once SongKong has finished a task such as Fix Songs it now automatically closes the progress dialog, previously it had to be manually dismissed.

SONGKONG-375:If user has disabled Discogs searching AND Matching dont show Unmatched Discogs section

The dialogs have been tidied up so that they do not show Discogs counters when the customer is not trying to match songs to Discogs.

Discogs Artists that could be linked to MusicBrainz Artists Report

Within MusicBrainz it is possible to link an artist to the equivalent Discogs artist, in fact there are already some 300,000 artists linked. However there are nearly a million artists within MusicBrainz leaving nearly 700,000 artists unlinked, the majority of these already existing in Discogs.

Linking both artists together allow information between the two databases to be shared, for example the Discogs artist may contain the real name of the artist and that might be missing from the equivalent MusicBrainz artist aliases. Alternatively MusicBrainz may have various translations of an artists names for different languages, and these variations may not be available in Discogs.

Albunack now contains a number of reports that attempt to find potential matches between the two together with a link for each match to make it easy to add a link from MusicBrainz artist to the Discogs artist, all listed here

The first report  MB artist matches Discogs artist and both have a release with same name and the artist names are unique contains over 40,000 potential links. Matches are only listed here if the following criteria are met:

• Simplified MusicBrainz artist is unique within MusicBrainz list of artists.
• Smplified Discogs artist name is unique within Discogs link of artists.
• Neither name is already matched
• Both the MB artist and the Discogs artist have released an album with the same name

Because of these conditions we are confident there are very few (if any at all) bad links.

The simplified artist names remove case, accents and some punctuation in order to allow matches between artist names that are not identical. Because we only include artists whose simplified name is unique within the databases it is unlikely that the MusicBrainz artist has linked to the wrong Discogs artist. And this is made very unlikely with the last condition that requires both artists to have a release with the same name.

Some of the links may already have already been created because the underlying database is not updated in real time, but it is regularly updated - usually every week.

Links are listed so that artists with the most releases are listed first, the basis being that these are likely to be of most interest.

But with 40,000 artists to link the job to complete this task could take some time, so if you have 10 minutes please consider double-checking and then creating some of these links within MusicBrainz.

Linking a Discogs release to a MusicBrainz Release

Linking a Discogs release to a MusicBrainz Release

In the past there have been two significant problems with trying to link Discogs releases to MusicBrainz release.

Finding a release

Firstly, it was difficult to Discogs release that can be linked to MusicBrainz but aren't already. You could browse an artist releases on Discogs but there is no indication on the Discogs site as to which ones are already linked to a MusicBrainz release. You can check whether an individual release on MusicBrainz is linked to the Discogs but you cannot filter a list of releases by whether or not they are already linked. 

And when popular artists can have hundreds of Discogs releases linked to a single master, and even moderately popular artists can often have more than thirty it is difficult to find the correct release to link to if any. Working out the best match is difficult when having to consider and compare differences and matches in metadata such as track lists, catalogue nos, barcodes, country of release and record labels.

Linking the Release

Secondly you have to actually add the link. 

Linking the release is not so difficult but when the Discogs release contains extra information that the MusicBrainz release does not have, wouldn't it be good to get that data in as well, but adding that extra data for more than a few releases at a time is time-consuming.

Linking a Discogs release to a MusicBrainz Release with Albunack

Today I am going to show you how to easily link Discogs releases to MusicBrainz releases with Albunacks artist-centric approach solves both of these problem

Finding a Release

Albunack combines data from both databases to provide comprehensive combined Artist discographies.:

• Information from MusicBrainz and Discogs is combined on a single page.
• Release group, Release and Track listings are all available on a single page.
• Existing links between MusicBrainz and Discogs are shown
• Potential matches identified by Albunack are shown. 

Linking a Release

When you click on the link button Albunack not only links the two releases but adds additional data such as barcodes and catalogue nos, you simply have to check the data and then click submit.

We have created a video to show exactly how this is done

Introductory video added to Albunack

The JThink Music Server is a dedicated server that contains MusicBrainz and Discogs data, and  is what Jaikoz and SongKong use for their automatic matching. Albunack is the public face of the music server and can be used to browse combined MusicBrainz/Discogs discographies for almost any artist.

If SongKong is not finding a match when you think it should, it is probably easier to check on Albunack instead of MusicBrainz or Discogs. Although note Albunack doesn't currently show collaborations or Various Artist releases (although they are in the JThink Music Server).

Two months ago we released Albunack and now we have now added  a Youtube channel , and this includes an introduction to Albunack



this gives a nice overview on how to navigate around an Albunack discography.

SongKong 3.15 Released

SongKong 3.15 Released

Today we release a new version of SongKong

Firstly this includes modifications required to work with cover art newly added to Discogs in the last two months, this is required due to a change in how Discogs provides their cover art data.

Then we have fixed a problem matching song title and track total when matching multi-disc Discogs releases.

We recommend you update to this latest version of SongKong and then run Fix Songs over your collection with  the For Songs Already Matched to MusicBrainz releases option set to Update Metadata and Filename Only. This will fix any data that may have been incorrectly added in earlier releases and will also updated your previously matched songs the very latest data added to MusicBrainz and Discogs.

New versions of Jaikoz and Jthink Music Server

New version of JThink Music Server 

Yesterday we released a new version of the JThink Music Server - this is what powers Jaikoz, SongKong and Albunack . This latest release fixes an issue with incorrect values for the Sort Artist field. It also add supports for the latest Discogs cover art, required because the way cover art has been provided by Discogs for releases added over the last couple of months has changed.

New version of Jaikoz Music Tagger 8.2.3

Today we release a new version of Jaikoz.

Firstly this includes modifications required to work with cover art added to Discogs in the last two months.

Then we have fixed two problems related to song titles and track total when matching multi-disc Discogs releases.

Then we have fixed a regression whereby the Only Update Year if Earlier option for the Year field was ignored so this field could be overwritten with a later date.

We recommend you update to this latest version of Jaikoz and then run Update Metadata from MusicBrainz and Update Metadata from Discogs to fix any data that may have been incorrectly added in earlier releases and to benefit from the very latest data added to MusicBrainz and Discogs.

Jaikoz 8.2.2 Available

A new release of  Jaikoz with a couple of bug fixes.

Delete Duplicate key check should be case insensitive

Delete Duplicate key check should be case insensitive to protect against tools that write identifiers with uppercase letters.

Only Allow Match if All Songs In Grouping Match to One Album should be true

Only Allow Match if All Songs In Grouping Match to One Album is defaulting to false on new install it should default to true otherwise existing songs grouped by album could be split if Jaikoz can match the songs but not all to the same album.

Albunack, bringing Discogs and MusicBrainz together.

MusicBrainz is the defacto standard for Music, MusicBrainz Ids are used by many of the major players in online Music services such as Google the BBC and Spotify, and the MusicBrainz schema is a thing of beauty. But it has always been difficult getting the latest releases into MusicBrainz because there are no automated feeds from record companies and it mainly relies on the hard work of the many contributors.

Discogs is another online database, its not as well designed as MusicBrainz and its codebase is not open, but it has more artists and more releases and does make its database available monthly under a public domain license. Discogs has a vibrant marketplace, but to sell a record it has to be in the database, this is a good reason for users to contribute releases, and because buyers need to know exactly what they are purchasing sellers need to enter the details accurately.

For years I have wanted to something to make it easier for editors to get data into MusicBrainz from Discogs in some form or other but never seemed to have the time, but over the last month I have been working on a solution and now I have beta available as: http://albunack.net



The idea is you look up an artist and it shows you a combined MusicBrainz + Discogs discography, showing releasegroups/release and tracks on a single page.

Here is an example for the Pale Saints, a greatly under appreciated UK band from the early 90's


This makes it easier to see gaps in the MusicBrainz which could be filled by data from Discogs, and makes it easier to add that data:

• It shows any existing links between MusicBrainz and Discogs matches. 
• It also matches up Musicbrainz releases without a Discogs link to potential Discogs releases and lists these matches together with a Link button to seed a MusicBrainz Edit release. 
• It provides an Import button to seed a MusicBrainz Add Release from other Discogs releases not linked to Musicbrainz  

I have been testing it myself over the last few weeks and it is so much quicker and so much safer  then previous ways I have used of adding data. Ive also found it is very good at showing when there is incorrect data in either MusicBrainz or Discogs.

You can also use it to check automated tagging done by SongKong or Jaikoz since they use the same underlying database.

Currently it only shows releases where the artist is the only album artist credit (i.e we don't list collaborations, nor Various Artist releases) Of course, the user still has to make decisions about the data and check the data before submitting anything to MusicBrainz, and everything still goes through the edit system. You can use the bug tracker for raising any issues.

Matching improvements with SongKong 3.14

A new release of SongKong 3.14 with a handful of bug fixes and improvements.

Unable to load large folder containing songs directly within it

If customers have all their files in one folder and that folder is on the network then SongKong can timeout before loading all files.

A few months ago the  Database Unreturned Connection Timeout option was set to 30 minutes to protect against against hibernate connection being deadlocked and therefore unable to return, it was part of a two prong defence against processes getting stuck. The other was adding a limit of 30 minutes for processing of any grouping- the original implementation of this was a little buggy but is now working correctly so we are safe to remove the  Database Unreturned Connection Timeout and this fixes the issue.

Release Date comparison is incorrect

When deciding between similar MusicBrainz releases we take the release date into account, and score earlier release dates higher. At one point the server only stored a single release date (and country) for a release but can now store multiple release date/country combinations (known as Release Events) but the client code was not updated to use release events, now fixed.

If fingerprint is AQAAAA dont check it

If Acoustid generates the above fingerprint for a song it means it could not actually generate a valid fingerprint so we should not try and lookup and Acoustid Id for it.

Discogs matching threshold checks should be higher

We have to be stricter when matching Discogs releases than MusicBrainz because we can not use AcoustId as a crosscheck, so the matching threshold has been increased.

If matching takes more than 24 hours time taken in summary just shows remainder
Both report and logs dont show correct timezone

Both these issues are to do with the displays of dates, I took the opportunity to familiarize myself with the new Java 8 Date/Time api to fix this.

Better song matching with SongKong 3.13

A new release of SongKong is now available that  fixes these two issues.

Firstly it fixes a regression introduced in 3.11

When reusing existing Discogs Releases as candidates SongKong was not checking trackcount properly. The check to see if the there are more tracks than the number of songs to match which is not allowed if you have Only allow match if all tracks on album were matched is incorrect. We tried to fix this in SongKong 3.12 but made a mistake but it is now fixed.

If MetadataMatcher fails if it needlessly does Match Recording Only again

If song matching by metadata  fails to find a match it needlessy does Match Recording Only, and there is no point because this would have already been done during the Match by Folder stage, worse than that metadata matching can be called a number of times for the same group of songs. So wasted processing that can cause timeouts if extending the length of matching for a grouping over the timelimit and use additional amounts of memory.