Results so far
writing on Overleaf
https://www.overleaf.com/project/65cb4260ce319380eaee2fb0
tables
dataset related table
performance table
Issues
Evaluation algorithm fixed
the reason perplexity of NB and REMI doesn't match is that there is an error in calculating probability.
Encoding matches for tempo changes
performance of REMI in 7 features
- REMI with 4 features(type, beat, pitch, duration)
- REMI with 7 features(type, beat, chord, tempo, pitch, duration, velocity)
- CP
Listening Test
available on google form
chords & code // harmony with structure
Your article on Symbolic Music Encoding is very insightful, especially as it bridges the gap between music theory and computational modeling. For those looking to take their research further, I'd highly recommend checking out this resource https://www.phdresearchproposal.org/engineering-research-proposal-topics/ It provides a solid foundation for those developing technical frameworks or exploring interdisciplinary projects that merge music, AI, and engineering fields. It could serve as an excellent guide for structuring proposals, particularly when addressing the complexities of symbolic music processing and other related topics. Keep up the great work!