[NIT-song70 dataset] About the Score(notes) data

[Jung Choi]

Hello,

Could anyone tell me how I could get access to the 'musical score' data for open Sinsy Japanese Song dataset, if it's publically open?
(Or should I create the matching musical score with raw audio file?)

Regards,

Jung Choi.

[Jung Choi]

Just to clarify,
there are note informations on the full label table,
but I need the exact timing of notes(as written on a score) to get time deviation between the notes and the vocal signals.

Jung.

[KAZUHIRO NAKAMURA]

Hello,

The music score data is not public because of the copyright problem.
The time information written on the label are same as the note boundary written on the music score.

Best regards,
Kazuhiro Nakamura

[Jung Choi]

Dear Kazuhiro,

Thank you very much for your reply!

I just want to check one thing for sure.

I'm trying to get the deviation between note(from score) boundaries and phoneme(from audio) boundaries.
I had been a little bit confused because the FULL label files also has 'current phoneme identity' as one of their labels.
However, I found that MONO label files has only phoneme timing boundaries, so I thought that FULL label files are divided into note boundaries(like you told me).

So,
1) For note boundaries, I should use FULL label files.
2) For phoneme boundaries, I should use MONO label files.

Is this a right understanding?

Best regards,
Jung Choi.

[KAZUHIRO NAKAMURA]

Dear Jung,

So,
1) For note boundaries, I should use FULL label files.
2) For phoneme boundaries, I should use MONO label files.
Is this a right understanding?

You are right.

Best regards,
Kazuhiro Nakamura.