Improvements repetition of artists PLGen Pro

[nelson c]

Hello Dmitry, yesterday I found that PLGen Pro had created a play list with two consecutive tracks from the same artist. (although this set 120 minutes minimum).

The artist is called "Man?"

When I see what happened in the log, I notice a slight difference between the metadata of the songs:

In the first the artist is typed as "Mana" and the second as "Man?"

A matching test shows they are only 65% similar. (when it is the same character, only contains a tilde)
Same with "?, ?, ?, ?"

I think it must be regarded as the character without the accent.

 

[nelson c]

I was also thinking about how to improve the filter to avoid the repetition of tracks where multiple artists collaborate.

It occurred to me the following:
That may setear catacteres to consider as an additional artist, for example: "And, and, Ft."

Artist should then analyze the metadata field and find all the artists in the song, and thus would avoid completely the repetition of tracks where there are collaborations.

Example:
Madonna ft. Nicki Minaj - Bitch I'm Madonna

Artist 1: Madonna
Artist 2: Nicki Minaj

Neither of these files may be repeated by the set time.

I hope I was clear

Thanks,
Nelson.

 

[djsoft]

A matching test shows they are only 65% similar. (when it is the same character, only contains a tilde)
Same with "?, ?, ?, ?"

I think it must be regarded as the character without the accent.

You're correct. This will be corrected and artists will be compared without considering accents/umlauts.

I was also thinking about how to improve the filter to avoid the repetition of tracks where multiple artists collaborate.

It occurred to me the following:
That may setear catacteres to consider as an additional artist, for example: "And, and, Ft."

Artist should then analyze the metadata field and find all the artists in the song, and thus would avoid completely the repetition of tracks where there are collaborations.

Example:
Madonna ft. Nicki Minaj - Bitch I'm Madonna

Artist 1: Madonna
Artist 2: Nicki Minaj

Neither of these files may be repeated by the set time.

I hope I was clear

Something like this was supposed to be implemented several years ago, but real life test showed that this kind of parser is often mistaken. There's simply too many ways for 2 or more artists to be mentioned in the aertist tag:
Artist1 vs Artist2
Artist1, Artist2
Artist1 ft. Artist2
Artist1 feat Artist2
Artist1 - Title (feat. Artist2) - here the other artist is included in the title field
the list goes on and on...
the partsing is made more difficult as some tags include more info like Club Mix, Entended Mix etc in the Artist field.

So in the result the partial artist name matching was implemented, but that's still not perfect.

We can do a second atempt in creating such a parser, and even make it configurable: specify artist separators, specify words that need to be excluded etc...

I'll addit to the to-do list for future versions.

If you have any suggestions about it I'll be glad to hear

 

[nelson c]

I was also thinking about how to improve the filter to avoid the repetition of tracks where multiple artists collaborate.

It occurred to me the following:
That may setear catacteres to consider as an additional artist, for example: "And, and, Ft."

Artist should then analyze the metadata field and find all the artists in the song, and thus would avoid completely the repetition of tracks where there are collaborations.

Example:
Madonna ft. Nicki Minaj - Bitch I'm Madonna

Artist 1: Madonna
Artist 2: Nicki Minaj

Neither of these files may be repeated by the set time.

I hope I was clear

Something like this was supposed to be implemented several years ago, but real life test showed that this kind of parser is often mistaken. There's simply too many ways for 2 or more artists to be mentioned in the aertist tag:
Artist1 vs Artist2
Artist1, Artist2
Artist1 ft. Artist2
Artist1 feat Artist2
Artist1 - Title (feat. Artist2) - here the other artist is included in the title field
the list goes on and on...
the partsing is made more difficult as some tags include more info like Club Mix, Entended Mix etc in the Artist field.

So in the result the partial artist name matching was implemented, but that's still not perfect.

We can do a second atempt in creating such a parser, and even make it configurable: specify artist separators, specify words that need to be excluded etc...

I'll addit to the to-do list for future versions.

If you have any suggestions about it I'll be glad to hear

Perfect, I agree that it will not be 100% effective and there demaciadas convinations possible, but this would greatly improve.

If the person who schedules the music, always writes metadata in the same way, this problem is eliminated.
For example Artist1 Ft. Artist2 - Name of the song
Personally I do not like to see:
Artist 1 - Name of the song (Ft. Artist2.)
Then change the metadata to the above manner, but I recognize that it is a very tedious job and most broadcasters do not.


Make this configurable text to separate artists is almost essential, for example in the Spanish language, besides the "Ft.", used "y" (the "and" in English). The same should happen in other languages

 

[djsoft]

If the person who schedules the music, always writes metadata in the same way, this problem is eliminated.

Make this configurable text to separate artists is almost essential, for example in the Spanish language, besides the "Ft.", used "y" (the "and" in English). The same should happen in other languages

Yes, if everything is tagged properly, and there's a freedon to choose separators, this may work. We'll give it another try in RadioBOSS 5.5 or 5.6, thanks for the suggestion.