mtpearce
commented
2 years ago Firstly, I'm glad to hear that you've been successfully using IDyOM for modelling harmony. Your approach of mapping chords to integers is essentially what we have done in previous research.[1][2] However, genuine multiple viewpoint representation of harmony is in the pipeline.
Second, the kern import module doesn't yet recognise dynamics, ornaments or articulation, so this is not the route to go down. I'd suggest creating .lisp files, which can be imported as follows:
(idyom-db:import-data :lisp ...)
The format should be clear from the following example but let me know if you have any questions:
("185 chorale melodies harmonised by J.S. Bach." 96 60 ("chor-001" ((:ONSET 72) (:DELTAST 0) (:BIOI 72) (:DUR 24) (:CPITCH 73) (:MPITCH 42) (:ACCIDENTAL 1) (:KEYSIG 3) (:MODE 0) (:BARLENGTH 96) (:PULSES 4) (:PHRASE 1) (:VOICE 1) (:ORNAMENT NIL) (:COMMA NIL) (:VERTINT12 NIL) (:ARTICULATION NIL) (:DYN NIL)) ((:ONSET 96) (:DELTAST 0) (:BIOI 24) (:DUR 12) (:CPITCH 73) (:MPITCH 42) (:ACCIDENTAL 1) (:KEYSIG 3) (:MODE 0) (:BARLENGTH 96) (:PULSES 4) (:PHRASE 0) (:VOICE 1) (:ORNAMENT NIL) (:COMMA NIL) (:VERTINT12 NIL) (:ARTICULATION NIL) (:DYN NIL)) ... ((:ONSET 960) (:DELTAST 0) (:BIOI 24) (:DUR 72) (:CPITCH 69) (:MPITCH 40) (:ACCIDENTAL 0) (:KEYSIG 3) (:MODE 0) (:BARLENGTH 96) (:PULSES 4) (:PHRASE -1) (:VOICE 1) (:ORNAMENT NIL) (:COMMA NIL) (:VERTINT12 NIL) (:ARTICULATION NIL) (:DYN NIL))) ... ("chor-185" ((:ONSET 72) (:DELTAST 0) (:BIOI 72) (:DUR 24) (:CPITCH 65) (:MPITCH 38) (:ACCIDENTAL 0) (:KEYSIG -1) (:MODE 0) (:BARLENGTH 96) (:PULSES 4) (:PHRASE 1) (:VOICE 1) (:ORNAMENT NIL) (:COMMA NIL) (:VERTINT12 NIL) (:ARTICULATION NIL) (:DYN NIL)) ... ((:ONSET 816) (:DELTAST 0) (:BIOI 24) (:DUR 24) (:CPITCH 65) (:MPITCH 38) (:ACCIDENTAL 0) (:KEYSIG -1) (:MODE 0) (:BARLENGTH 96) (:PULSES 4) (:PHRASE -1) (:VOICE 1) (:ORNAMENT NIL) (:COMMA NIL) (:VERTINT12 NIL) (:ARTICULATION NIL) (:DYN NIL))))
[1] Sears, D., Pearce, M. T., Caplin, W. E., & McAdams, S. (2018). Simulating melodic and harmonic expectations for tonal cadences using probabilistic models http://webprojects.eecs.qmul.ac.uk/marcusp/papers/SearsEtAl2018.pdf. /Journal of New Music Research/, 47, 29-52. https://doi.org/10.1080/09298215.2017.1367010 [2] Cheung, V., Harrison, P. M. C., Meyer, L., Pearce, M. T., Haynes, J-D, & Koelsch, S. (2019). Uncertainty and surprise jointly predict musical pleasure and amygdala, hippocampus, and auditory cortex activity. http://webprojects.eecs.qmul.ac.uk/marcusp/papers/CheungEtAl2019.pdf /Current Biology/, 29(23), 4084-4092.e4. https://doi.org/10.1016/j.cub.2019.09.067
On 04/06/2022 20:52, adityac95 wrote:
I'm trying to get IDyOM to model harmonic structure. I know it's only designed for monophonic input right now, so in the past I've just been converting chord labels to an alphabet of integers, making .txt files of the pieces, and training IDyOM on those. That's been fine but it only has single viewpoint functionality. I recently came up with a representation of chords in my corpus of interest that I can essentially map to a purely monophonic input by leveraging some of the non-pitched viewpoints (dynamics, ornaments and articulation) in order to get the multiple viewpoints functionality working for me.
The corpus is currently in a tabular format and I'm going to convert it to a number of humdrum .krn files. The format has functionality for all of the viewpoints I listed above. However, I'm not sure of two things. First, I wanted to double check which humdrum symbols match each of the articulations that IDyOM can understand. For ease of reference, here they are:
articulation: 0 = no articulation mark; 1 = staccato; 2 = staccatissimo; 3 = sforzando; 4 = marcatoI'm particularly unsure about the sforzando since it's often treated as a dynamic marking.
Secondly, will IDyOM definitely know how to interpret a separate dynamics **dynam spine if it's in the .krn file? Or should I be encoding them in some other way to guarantee that they're read?
— Reply to this email directly, view it on GitHub https://github.com/mtpearce/idyom/issues/44, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACZS6JZW5CH6V7TP6NAJ6JLVNOXZHANCNFSM5X35CAJQ. You are receiving this because you are subscribed to this thread.Message ID: @.***>
-- School of Electronic Engineering and Computer Science Queen Mary University of London, London E1 4NS, UK Web: http://webprojects.eecs.qmul.ac.uk/marcusp/ Lab: http://music-cognition.eecs.qmul.ac.uk/
adityac95
I'm trying to get IDyOM to model harmonic structure. I know it's only designed for monophonic input right now, so in the past I've just been converting chord labels to an alphabet of integers, making .txt files of the pieces, and training IDyOM on those. That's been fine but it only has single viewpoint functionality. I recently came up with a representation of chords in my corpus of interest that I can essentially map to a purely monophonic input by leveraging some of the non-pitched viewpoints (dynamics, ornaments and articulation) in order to get the multiple viewpoints functionality working for me.
The corpus is currently in a tabular format and I'm going to convert it to a number of humdrum .krn files. The humdrum format has functionality for all of the viewpoints I listed above. However, I'm not sure of two things. First, I wanted to double check which humdrum symbols match each of the articulations that IDyOM can understand. For ease of reference, here they are:
I'm particularly unsure about the sforzando since it's often treated as a dynamic marking.
Secondly, will IDyOM definitely know how to interpret a separate dynamics **dynam spine if it's in the .krn file? Or should I be encoding them in some other way to guarantee that they're read?