9 ms·

I’ve always wondered, is there a better way to annotate music? Essentially, we are transcribing temporal patterns into spatial 2D patterns. Discrete notes were

I’ve always wondered, is there a better way to annotate music? Essentially, we are transcribing temporal patterns into spatial 2D patterns. Discrete notes were created to allow for hand written notation, but if we assume notation doesn’t need to be handwriting-compatible, I wonder if we can do a better job of capturing details? Perhaps some kind of spectral/FFT scroll?

Mediterraneo10 8y ago
In the late 20th-century, as tape parts and live electronics became more common in avant-garde classical music, various novel conventions arose to represent the electronic part on the score. This was not so people could duplicate the electronic parts themselves – they had often been painstakingly created in centers like IRCAM over months, and so they were provided to performers on a tape or as a collection of software. But the representation of the electronic part on the score helped the human performers playing their traditionally-noted parts keep in time with the electronics.
mhh__ 8y ago
Yes and no. You need to be able to sight read it, so it has to be similar to the current system, but without the baggage.
xtagon 8y ago
I've wondered the same thing. Guitar tablature is a more intuitive notation for me to understand than piano, by far, so I wonder if there is an alternative for piano that feels as simple. There are some[0], but not quite as radical as I would expect, they still seem to use the same basic look and feel. [0] http://musicnotation.org/systems/ http://musicnotation.org/systems/
- 0815test 8y ago
  Guitar tablature is not a notation for music though, it's a notation for how-to-play-something-on-the-guitar. And yes, that's a very relevant distinction here - when it comes to music, there's nothing more intuitive than the well-established notations that we already have (with a few very limited caveats, of course). And I'm not just talking about Western notation, either - the number-based notation that's in widespread use in East Asia is also highly intuitive and perhaps more so.
  - baddox 8y ago
    And that distinction really only exists for instruments like the guitar that can play the same note on multiple strings. Piano music on staves is already very similar to a piano roll, at least on the y axis. You still have to learn the notation for different durations (the x axis), but there's not much mystery over how to physically play the notation on a piano, other than some potential for tricky fingerings. There is also fairly standard staff notation for fingerings, both for piano music and classical guitar music (in my experience, it's much more important for the latter).
  - caillancm 8y ago
    >And I'm not just talking about Western notation, either - the number-based notation that's in widespread use in East Asia is also highly intuitive and perhaps more so. Do you have any more information about this? Fpr example, scores of Chinese music using the Indian/Arabic number system? I know that number notation is used as a mataphor for gamelan music—by westerers, which has caught on natively (although which might be frowned upon) for archival purposes, at least. The number metaphor for tones clashes with the number metapor for the counting of beats, which is ubiquitous in Western music. But perhaps there are different metaphores in East-Asian music.
- Yetanfou 8y ago
  There is Klavarskribo [1], more or less the keyboard-equivalent of Tabulature for stringed instruments. [1] https://en.wikipedia.org/wiki/Klavarskribo https://en.wikipedia.org/wiki/Klavarskribo
  - xtagon 8y ago
    I like this! It's sort of like the piano rolls you see in DAW sequencers.
    - zserge 8y ago
      I've made a toy page for it - https://naivesound.com/klavar/ https://naivesound.com/klavar/ Click on the columns to put a note, add more height if needed, edit the title, then print it to PDF. I often use it to write down simple melodies when playing the piano. Also, there are two blank PDF templates - https://github.com/naivesound/naivesound.github.io/tree/master/klavar https://github.com/naivesound/naivesound.github.io/tree/mast...
handedness 8y ago
Relevant: https://news.ycombinator.com/item?id=17993223 https://news.ycombinator.com/item?id=17993223
adrianh 8y ago
Our product Soundslice is more or less trying to do this: https://www.soundslice.com/ https://www.soundslice.com/ The philosophy is that music notation will never be able to capture the subtlety of a performance — so syncing the notation with real recordings/videos is a much richer learning experience than notation alone.
- beat 8y ago
  Soundslice is pretty awesome!
- mhh__ 8y ago
  Useful but if it's intended for learning (to play not to play), a systematic way of describing the harmony would be a good step forward
beat 8y ago
The map is not the territory. Furthermore, the language of the map limits how the territory can be described. So "better way to annotate music" depends, at least in part, on the kind of music you're annotating. This really came home to me when I was studying frame drumming of the central Asian plains. In Western notation terms, this music has very complex and difficult "time signatures", but that's because it's not metric in the same way as European music. If you think of it in its own terms, it's actually pretty easy to understand. Just break the rhythm down into sub-beats of 2, 3, or 4 pulses. So for example, a 13 beat long line might be 23233. Try counting that out loud - one two one two three one two one two three one two three - and you'll immediately hear it. But it's hell to notate in Western notation. And it gets worse when pieces turn into meta-sequences that have "bars" of varying length. I've long been involved with the free improvisation scene, and improvisation always happens within some sort of set of rules. How do you notate something that's made up on the spot, within a framework? Take blues guitar leads as an example. They tend to stick within pretty limited pentatonic scales, against the underlying triplet shuffle feel of American music, but they're not "notated", really. This leads to other, alternative notation systems that more aptly describe the style of music at hand, like the Nashville Number System.
- microtherion 8y ago
  Reminds me of a Charles Mingus anecdote I came across today: > The reed player Yusef Lateef recalled his own experience learning Mingus music: “On one composition I had a solo and, as opposed to having chord symbols for me to improvise against, he had drawn a picture of a coffin. And that was the substance upon which I was to improvise.” https://fdleone.com/2016/01/29/charles-mingus-the-jazz-workshop-concerts-1964-1965/ https://fdleone.com/2016/01/29/charles-mingus-the-jazz-works...
- 0815test 8y ago
  Additive meter is actually a well-known concept even in Western music. It's clearly needed even to make sense of, e.g. "5/4" or "7/4" meter, which yes, are sometimes found in the repertoire - because these are already ambiguous in practice: does the 5 in "5/4" mean "3+2" or "2+3"? You need to know that in order to understand the music and figure out how it goes - to make it "meaningful" as someone else said.
jdietrich 8y ago
>I’ve always wondered, is there a better way to annotate music? Not really. Western staff notation is heavily optimised for intelligibility by an expert user. It's not very intuitive, it barely works at all for non-western pitch systems, but a trained musician can read, understand and reproduce complex scores in real time. For the primary users of western musical notation, that is an essential, non-negotiable feature that negates every disadvantage. Western notation leans heavily on both visual pattern recognition and an understanding of music theory. It can be read in chunks, like written natural language. When you're reading this comment, you're not parsing individual letters, you're recognising whole words and phrases. It's the same for music notation - a skilled user sees chords, progressions and scale fragments rather than individual notes. They are parsing the dots and lines as meaningful musical information, not purely abstract pitch and duration data. It's not a particularly precise or machine-readable system, but it's not supposed to be. Where we need machine-readability, we have MIDI and OSC - they're totally unreadable by humans, but they convey a vastly greater range of data at vastly greater precision than any traditional notation system. Notation encodes as much detail as is musically relevant and no more. For example, we typically only use eight levels of loudness, from ppp to fff. It's not really useful for the composer or the performer to encode very precise loudness information, because human musicians can't reproduce that level of precision.
- tropo 8y ago
  We could have that same intelligibility for less-expert users if we didn't have all the legacy gunk. Of course, the expert users have a vested interest in having a more-exclusive club. Nearly all of them would reject anything new. A simple improvement is to give each of the 12 common pitches within the octave a distinct position on the staff. This eliminates accidentals from typical music, leaving them only needed for quarter-tone compositions. Sight-transposition becomes trivial, possibly within reach of beginners.
  - 0815test 8y ago
    > A simple improvement is to give each of the 12 common pitches within the octave a distinct position on the staff. This doesn't work - the staff becomes too large, or the registral range it covers becomes way too small. The current staff has evolved to be readable without expressly counting the lines or spaces, via visual subitizing. If you add more lines, you lose this property and sight-reading becomes infeasible. Plus, diatonic pitch is actually meaningful in a musical sense, and it's useful to have a notation that references it. That's why we use for instance C, D, E, F, G and not <0>, <2>, <4>, <5>, <7>. (This also describes the very real problems w/ Klavarskribo, BTW.)
archagon 8y ago
I played around a bit with stroke-based notation a few years back: http://archagon.net/blog/2016/02/05/rethinking-musical-notation-with-composers-sketchpad/ http://archagon.net/blog/2016/02/05/rethinking-musical-notat... In theory, you could annotate arbitrary pitch-based music with this approach, even if it uses different tunings or scales. If I have more time, I'd like to develop it into a more robust product one of these days—especially with things like MIDI Polyphonic Expression gaining traction.
Riverheart 8y ago
Here's one attempt. https://www.hummingbirdnotation.com https://www.hummingbirdnotation.com
- bwest87 8y ago
  Thanks for posting. I'm the creator of this notation. If anyone has any questions, I'm happy to give my thoughts!
holy_city 8y ago
You can perfectly represent music using a 1D representation, or 2D, it doesn't really matter because Musical notation is a cognitive problem, not a technical one (the latter was solved by Shannon). It's not so much how do you encode pitches and timing information (this is trivial today), but how to convey the meaning of the encoding and how effectively a musician can recreate the piece from the representation. Weaver pointed this in his forward to Shannon's magnum opus [1], on the topic of the three levels problems in communication. The first level, the "technical problems" are those of encoding information into symbols for transmission (in the abstract sense), whereas the "semantic" problems cover how one conveys intent or meaning behind the information (and that it is impossible for meaning to be clarified by the sender, in general), and finally the "effectiveness" problems which cover how the receiver conducts action based on the information communicated. Musical notation is purely a semantic and effectiveness problem, and over the past few thousand years we have found that there is no optimal solution, and that listening to the original music and having direction or instruction are as important as the symbols on the page. The notation itself is a simplex transmission from the author (or transcriber) to the performer, instruction/direction helps make it somewhat duplex to improve (but not solve) the semantic problem, while listening is the end-all, be-all solution provided you have source material recorded and training to interpret it. The forward to Shannon's theory of communication is really essential reading for all software/electrical engineers imo: [1] The Mathematical Theory of Communication, Weaver & Shannon http://www.magmamater.cl/MatheComm.pdf http://www.magmamater.cl/MatheComm.pdf
- nixpulvis 8y ago
  I really, REALLY like these 3 "levels". They map to programming language perfectly (as we'd expect), syntax is encoding, semantics is the logic we think we're going to get (hypothesis, remember it's computer "science"), and the effectiveness is what actually happens. We love computers because they almost always do what we want them to do, but people like to forget that nothing is perfect. All communication, regardless of with whom or what, is lossy.
- TheOtherHobbes 8y ago
  A piece of classical music is not a static data structure that can be compressed and sent over a channel. Classical musicians don't recreate pieces, they actively create them. The fact that scores allow different interpretations is a feature, not a bug. There is no definitive interpretation, and there is no "semantic problem" to "solve." And there is a lot more involved than playing the right notes at the right times in the right order. Not even a performance by the composer is definitive.
anigbrowl 8y ago
Yes, there's lots of different approaches (I've been fiddling with these ideas for 20 years). Among the most interesting is conversion of note data to polar coordinates, using pitch class as the angular component and octave as as the amplitude. You can also represent musical pitches on hexagonal layouts, as graph structures, perturbations of strange attractors in n-d phase space...
- caillancm 8y ago
  >You can also represent musical pitches on hexagonal layouts, as graph structures, perturbations of strange attractors in n-d phase space... Interesting conecpts. Are there any examples you could post?
brilee 8y ago
I wrote a blog post trying to explain some of the subtleties of Western musical notation - http://www.moderndescartes.com/essays/musical_notation/ http://www.moderndescartes.com/essays/musical_notation/ I think better systems exist, but it's hard to design one without understanding the span of functionality that the current system is capable of handling. And once you've spent the 15 years it takes to fluently read Western musical notation, it's hard to think any other way...

Mediterraneo10 8y ago

In the late 20th-century, as tape parts and live electronics became more common in avant-garde classical music, various novel conventions arose to represent the electronic part on the score. This was not so people could duplicate the electronic parts themselves – they had often been painstakingly created in centers like IRCAM over months, and so they were provided to performers on a tape or as a collection of software. But the representation of the electronic part on the score helped the human performers playing their traditionally-noted parts keep in time with the electronics.

mhh__ 8y ago

Yes and no. You need to be able to sight read it, so it has to be similar to the current system, but without the baggage.

xtagon 8y ago

I've wondered the same thing. Guitar tablature is a more intuitive notation for me to understand than piano, by far, so I wonder if there is an alternative for piano that feels as simple. There are some[0], but not quite as radical as I would expect, they still seem to use the same basic look and feel. [0] http://musicnotation.org/systems/ http://musicnotation.org/systems/

0815test 8y ago
Guitar tablature is not a notation for music though, it's a notation for how-to-play-something-on-the-guitar. And yes, that's a very relevant distinction here - when it comes to music, there's nothing more intuitive than the well-established notations that we already have (with a few very limited caveats, of course). And I'm not just talking about Western notation, either - the number-based notation that's in widespread use in East Asia is also highly intuitive and perhaps more so.
- baddox 8y ago
  And that distinction really only exists for instruments like the guitar that can play the same note on multiple strings. Piano music on staves is already very similar to a piano roll, at least on the y axis. You still have to learn the notation for different durations (the x axis), but there's not much mystery over how to physically play the notation on a piano, other than some potential for tricky fingerings. There is also fairly standard staff notation for fingerings, both for piano music and classical guitar music (in my experience, it's much more important for the latter).
- caillancm 8y ago
  >And I'm not just talking about Western notation, either - the number-based notation that's in widespread use in East Asia is also highly intuitive and perhaps more so. Do you have any more information about this? Fpr example, scores of Chinese music using the Indian/Arabic number system? I know that number notation is used as a mataphor for gamelan music—by westerers, which has caught on natively (although which might be frowned upon) for archival purposes, at least. The number metaphor for tones clashes with the number metapor for the counting of beats, which is ubiquitous in Western music. But perhaps there are different metaphores in East-Asian music.
Yetanfou 8y ago
There is Klavarskribo [1], more or less the keyboard-equivalent of Tabulature for stringed instruments. [1] https://en.wikipedia.org/wiki/Klavarskribo https://en.wikipedia.org/wiki/Klavarskribo
- xtagon 8y ago
  I like this! It's sort of like the piano rolls you see in DAW sequencers.
  - zserge 8y ago
    I've made a toy page for it - https://naivesound.com/klavar/ https://naivesound.com/klavar/ Click on the columns to put a note, add more height if needed, edit the title, then print it to PDF. I often use it to write down simple melodies when playing the piano. Also, there are two blank PDF templates - https://github.com/naivesound/naivesound.github.io/tree/master/klavar https://github.com/naivesound/naivesound.github.io/tree/mast...

handedness 8y ago

Relevant: https://news.ycombinator.com/item?id=17993223 https://news.ycombinator.com/item?id=17993223

adrianh 8y ago

Our product Soundslice is more or less trying to do this: https://www.soundslice.com/ https://www.soundslice.com/ The philosophy is that music notation will never be able to capture the subtlety of a performance — so syncing the notation with real recordings/videos is a much richer learning experience than notation alone.

beat 8y ago
Soundslice is pretty awesome!
mhh__ 8y ago
Useful but if it's intended for learning (to play not to play), a systematic way of describing the harmony would be a good step forward

beat 8y ago

The map is not the territory. Furthermore, the language of the map limits how the territory can be described. So "better way to annotate music" depends, at least in part, on the kind of music you're annotating. This really came home to me when I was studying frame drumming of the central Asian plains. In Western notation terms, this music has very complex and difficult "time signatures", but that's because it's not metric in the same way as European music. If you think of it in its own terms, it's actually pretty easy to understand. Just break the rhythm down into sub-beats of 2, 3, or 4 pulses. So for example, a 13 beat long line might be 23233. Try counting that out loud - one two one two three one two one two three one two three - and you'll immediately hear it. But it's hell to notate in Western notation. And it gets worse when pieces turn into meta-sequences that have "bars" of varying length. I've long been involved with the free improvisation scene, and improvisation always happens within some sort of set of rules. How do you notate something that's made up on the spot, within a framework? Take blues guitar leads as an example. They tend to stick within pretty limited pentatonic scales, against the underlying triplet shuffle feel of American music, but they're not "notated", really. This leads to other, alternative notation systems that more aptly describe the style of music at hand, like the Nashville Number System.

microtherion 8y ago
Reminds me of a Charles Mingus anecdote I came across today: > The reed player Yusef Lateef recalled his own experience learning Mingus music: “On one composition I had a solo and, as opposed to having chord symbols for me to improvise against, he had drawn a picture of a coffin. And that was the substance upon which I was to improvise.” https://fdleone.com/2016/01/29/charles-mingus-the-jazz-workshop-concerts-1964-1965/ https://fdleone.com/2016/01/29/charles-mingus-the-jazz-works...
0815test 8y ago
Additive meter is actually a well-known concept even in Western music. It's clearly needed even to make sense of, e.g. "5/4" or "7/4" meter, which yes, are sometimes found in the repertoire - because these are already ambiguous in practice: does the 5 in "5/4" mean "3+2" or "2+3"? You need to know that in order to understand the music and figure out how it goes - to make it "meaningful" as someone else said.

jdietrich 8y ago

>I’ve always wondered, is there a better way to annotate music? Not really. Western staff notation is heavily optimised for intelligibility by an expert user. It's not very intuitive, it barely works at all for non-western pitch systems, but a trained musician can read, understand and reproduce complex scores in real time. For the primary users of western musical notation, that is an essential, non-negotiable feature that negates every disadvantage. Western notation leans heavily on both visual pattern recognition and an understanding of music theory. It can be read in chunks, like written natural language. When you're reading this comment, you're not parsing individual letters, you're recognising whole words and phrases. It's the same for music notation - a skilled user sees chords, progressions and scale fragments rather than individual notes. They are parsing the dots and lines as meaningful musical information, not purely abstract pitch and duration data. It's not a particularly precise or machine-readable system, but it's not supposed to be. Where we need machine-readability, we have MIDI and OSC - they're totally unreadable by humans, but they convey a vastly greater range of data at vastly greater precision than any traditional notation system. Notation encodes as much detail as is musically relevant and no more. For example, we typically only use eight levels of loudness, from ppp to fff. It's not really useful for the composer or the performer to encode very precise loudness information, because human musicians can't reproduce that level of precision.

tropo 8y ago
We could have that same intelligibility for less-expert users if we didn't have all the legacy gunk. Of course, the expert users have a vested interest in having a more-exclusive club. Nearly all of them would reject anything new. A simple improvement is to give each of the 12 common pitches within the octave a distinct position on the staff. This eliminates accidentals from typical music, leaving them only needed for quarter-tone compositions. Sight-transposition becomes trivial, possibly within reach of beginners.
- 0815test 8y ago
  > A simple improvement is to give each of the 12 common pitches within the octave a distinct position on the staff. This doesn't work - the staff becomes too large, or the registral range it covers becomes way too small. The current staff has evolved to be readable without expressly counting the lines or spaces, via visual subitizing. If you add more lines, you lose this property and sight-reading becomes infeasible. Plus, diatonic pitch is actually meaningful in a musical sense, and it's useful to have a notation that references it. That's why we use for instance C, D, E, F, G and not <0>, <2>, <4>, <5>, <7>. (This also describes the very real problems w/ Klavarskribo, BTW.)

archagon 8y ago

I played around a bit with stroke-based notation a few years back: http://archagon.net/blog/2016/02/05/rethinking-musical-notation-with-composers-sketchpad/ http://archagon.net/blog/2016/02/05/rethinking-musical-notat... In theory, you could annotate arbitrary pitch-based music with this approach, even if it uses different tunings or scales. If I have more time, I'd like to develop it into a more robust product one of these days—especially with things like MIDI Polyphonic Expression gaining traction.

Riverheart 8y ago

Here's one attempt. https://www.hummingbirdnotation.com https://www.hummingbirdnotation.com

bwest87 8y ago
Thanks for posting. I'm the creator of this notation. If anyone has any questions, I'm happy to give my thoughts!

holy_city 8y ago

You can perfectly represent music using a 1D representation, or 2D, it doesn't really matter because Musical notation is a cognitive problem, not a technical one (the latter was solved by Shannon). It's not so much how do you encode pitches and timing information (this is trivial today), but how to convey the meaning of the encoding and how effectively a musician can recreate the piece from the representation. Weaver pointed this in his forward to Shannon's magnum opus [1], on the topic of the three levels problems in communication. The first level, the "technical problems" are those of encoding information into symbols for transmission (in the abstract sense), whereas the "semantic" problems cover how one conveys intent or meaning behind the information (and that it is impossible for meaning to be clarified by the sender, in general), and finally the "effectiveness" problems which cover how the receiver conducts action based on the information communicated. Musical notation is purely a semantic and effectiveness problem, and over the past few thousand years we have found that there is no optimal solution, and that listening to the original music and having direction or instruction are as important as the symbols on the page. The notation itself is a simplex transmission from the author (or transcriber) to the performer, instruction/direction helps make it somewhat duplex to improve (but not solve) the semantic problem, while listening is the end-all, be-all solution provided you have source material recorded and training to interpret it. The forward to Shannon's theory of communication is really essential reading for all software/electrical engineers imo: [1] The Mathematical Theory of Communication, Weaver & Shannon http://www.magmamater.cl/MatheComm.pdf http://www.magmamater.cl/MatheComm.pdf

nixpulvis 8y ago
I really, REALLY like these 3 "levels". They map to programming language perfectly (as we'd expect), syntax is encoding, semantics is the logic we think we're going to get (hypothesis, remember it's computer "science"), and the effectiveness is what actually happens. We love computers because they almost always do what we want them to do, but people like to forget that nothing is perfect. All communication, regardless of with whom or what, is lossy.
TheOtherHobbes 8y ago
A piece of classical music is not a static data structure that can be compressed and sent over a channel. Classical musicians don't recreate pieces, they actively create them. The fact that scores allow different interpretations is a feature, not a bug. There is no definitive interpretation, and there is no "semantic problem" to "solve." And there is a lot more involved than playing the right notes at the right times in the right order. Not even a performance by the composer is definitive.

anigbrowl 8y ago

Yes, there's lots of different approaches (I've been fiddling with these ideas for 20 years). Among the most interesting is conversion of note data to polar coordinates, using pitch class as the angular component and octave as as the amplitude. You can also represent musical pitches on hexagonal layouts, as graph structures, perturbations of strange attractors in n-d phase space...

caillancm 8y ago
>You can also represent musical pitches on hexagonal layouts, as graph structures, perturbations of strange attractors in n-d phase space... Interesting conecpts. Are there any examples you could post?

brilee 8y ago

I wrote a blog post trying to explain some of the subtleties of Western musical notation - http://www.moderndescartes.com/essays/musical_notation/ http://www.moderndescartes.com/essays/musical_notation/ I think better systems exist, but it's hard to design one without understanding the span of functionality that the current system is capable of handling. And once you've spent the 15 years it takes to fluently read Western musical notation, it's hard to think any other way...