Closed asigalov61 closed 5 months ago
Hi Alex,
Definitely interested. Perhaps it would be easy to add lyrics as a feature in MidiCaps for those available. Happy to discuss more on collaboration, perhaps we can discuss over email.
Sent from my mobile device.
On Wed, 26 Jun 2024 at 19:56, Alex @.***> wrote:
Hey AMAAI team :)
First of all, I wanted to thank you for creating and sharing this MidiCaps dataset! Its very nice and will be very useful for my projects :)
Secondly, I wanted to ask you if you would consider incorporating description/captions based on songs lyrics? LAKH MIDI Dataset contains a lot of songs for which there are lyrics available.
Incorporating lyrics captions/summaries would allow creation of the models or search/recommendation systems that would be based on the song description which would produce better results.
I have created and tested such dataset and system/model here: https://huggingface.co/spaces/asigalov61/Music-Sentence-Transformer
Check it out! I think it may give you some new ideas and/or inspiration :)
I also would be happy to colab on something like that if you are interested.
Sincerely,
Alex
— Reply to this email directly, view it on GitHub https://github.com/AMAAI-Lab/MidiCaps/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEES66AYOLEW7QF2AFYENLZJKT5FAVCNFSM6AAAAABJ5XLGR2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3TKMJTGEYTMMI . You are receiving this because you are subscribed to this thread.Message ID: @.***>
@dorienh Thank you for your response :)
I will email you shortly :)
Alex.
Hey AMAAI team :)
First of all, I wanted to thank you for creating and sharing this MidiCaps dataset! Its very nice and will be very useful for my projects :)
Secondly, I wanted to ask you if you would consider incorporating description/captions based on songs lyrics? LAKH MIDI Dataset contains a lot of songs for which there are lyrics available.
Incorporating lyrics captions/summaries would allow creation of the models or search/recommendation systems that would be based on the song description which would produce better results.
I have created and tested such dataset and system/model here: https://huggingface.co/spaces/asigalov61/Music-Sentence-Transformer
Check it out! I think it may give you some new ideas and/or inspiration :)
I also would be happy to colab on something like that if you are interested.
Sincerely,
Alex