“Using Case Based Reasoning.”
Go here. If you want to, that is.
Something vaguely wrong about this. “Open the pod bay doors, Hal. Give us a tune while you’re at it.”
I was hoping for more in his formal description of Irish music, but it’s shallowly general.
I’d be quite interested in seeing a statistical analysis of tune form from the abc archive.
Maybe Duggan thinks there just aren’t enough tunes. Plug in an algorithm, and welcome to the Brave New World.
There’s a rather important word missing from his keyword list: soul.
It’s an interesting idea. I don’t think his description of Irish music is all that important in the context of what he is trying to do. He made a reasonable stab at it in an introduction to what is essentially a computer-centred project. I hope he succeeds in his endeavour if only to demonstrate whether a tune that would stand the test of time could be created using case based reasoning. I doubt it personally but it is an interesting idea worth persuing. Some tunes you hear or even get out of collections are so melodically banal they could well have been generated using case based reasoning.
Maybe if he threw a few Brendan Ring, Mike McGoldrick, Diarmuid Moynihan and John Carthy tunes into the mix he would come up with something interesting…
Steve, what is this thing “soul” of which you speak?
Ah, my friend. 'Tis the very quintessence, the true flavour, the still yet moving light that we can reach out for yet not touch, that which lies at the very heart of things. The old man with the squawky fiddle scratching out a solemn yet strangely gay reel. Tommy Peoples playing Lucy Campbell, his eyes cast slightly down and not seeing, yet all-seeing. An ancient recording of James Morrison, awful to listen to and peppered with mistakes, but somehow, oddly, utterly divine. The raptness of listening to a well-played slow air, so good that you scarcely dare breathe from its start to its finish.
Am I getting there? 
I think what you are talking about is clearly that ethereal characteristic which Tony McMahon referred to as “the cry of the curlew in the mountains” (when he bemoaned it’s absence in the playing of concertina virtuoso Niall Vallely). Unfortunately for Tony the curlew happens to be a wading seabird which explains it’s absence from the mountains and makes his point a bit pointless. 
Now Tommy Peoples playing Lucy Campbell is another matter. That is called technique. 
Damn it, man, don’t belabor poetry with the facts.
You are both wrong and Tony McMahon is fully exonerated. The curlew is indeed a wader but its lonely, melancholy cry is as likely to to be heard in the mountains as it is on the broad expanses of estuaries. I hear it regularly on Bude Marshes nature reserve as well as high up in the fastnesses of Bodmin Moor. You may even hear it over cities during migration flights. You have to be careful though. Around here, the curlew’s call is often skilfully mimicked by that annoying but exceedingly clever scoundrel, the starling.
Ah, but one thing Tommy Peoples and I have in common is that we can both play Lucy Campbell. Thing is, who would you rather hear play it? And why? If the answer to the second question is merely “technique,” I think you may be missing something. 
Now, bear in mind Tony McMahon was actually quoting John Kelly there. John was prone to the odd malapropism. Paul de Grae in a recent communication elsewhere said:
As well as being a magnificent musician, John Kelly senior was the Mrs. Malaprop of Irish music (so I gather from stories I’ve heard–to my shame, I never met him, despite passing his door in Capel Street every day when I was attending college nearby, having little interest in choons in those far-off days). Once he was giving advice to one of the boys who was thinking of getting married.
“Marriage?” said he, “it’s no Ethiopia, you know!”
On another occasion, having sat too long in an uncomfortable chair while playing for dancers, he remarked, “I have a pain in my boldoon.” A puzzled member of the public asked him what a boldoon was, and John replied, “'Tis a very important part of a man’s autonomy.”
I think we can appreciate the intention of what was said rather than dismiss it altogether.
I recall discussing the merits and practicalities of doing something like this here on C&F a few years ago, but I lacked the technical now-how to try.
The theory was that if you collect enough tunes of a particular type (e.g. jigs) and load them into a database, you could then create “new tunes” starting from sections of other tunes.
You might let the computer do all the “composing”, or start from phrase 1 and have the computer “suggest” new phrases from which you choose the one you want next, or have the computer “suggest” a whole tune which you then develop from, etc.
How interesting, soulful or playable these computer generated tunes might be is of course a matter of opinion, but the fact that it is now being done is in itself interesting.
Hmm (spake the Luddite ), that’s an interesting thought. I’m no box player, but playing the harmonica I couldn’t help noticing, for example, how easily a lot of those Cork and Kerry slides and polkas, in both D and A (yeah, yeah, modes, I know), sit so easily on a simple diatonic harmonica in D. It’s clearly to do with the layout of the notes on the box. Beethoven wrote music for the voice that is distressingly difficult to execute properly. He was an unsympathetic chap. There’s one sustained passage of a high note in the finale of the Choral Symphony that has the sopranos fighting good and proper. But, whoever they were in the mists of time who had a hand in those great Sliabh Luachra tunes, they were a bit craftier than Ludwig. The tunes were made to fit the instruments available, and that fitting was best done by the players themselves I’ll be bound. Achieving that without compromising the wonderful innate musicality of the tunes is one of the glories of ITM. Now, a computer that could have the subconscious of a fiddle or flute player when composing tunes - that would be something. Still, even with rather mechanical tunes I suppose the usual evolution would take place once they were in the hands of the players. If you could actually fool anyone into playing them, that is!
My hunch is that your question is really more to point of the project. Here are his last two bullets under “Possible usage scenarios”:
- To better understand what makes a “good” tune or a “good” set
- To better understand the creative process in humans and machines.
One way to try to understand the creative process is to try to create something yourself, and then evaluate the results. In this case, the writer is testing both his analysis scheme of the existing ITM tunes, and his algorithm for generating more of them by generating tunes, and then asking people if they’re “good.”
To this non-expert, it looks a lot like a Turing test (http://www.turing.org.uk/turing/scrapbook/test.html) sort of situation.
Why ITM? Because it’s what Douglas Hofstadter calls a “microdomain”: an overtly simple structure that is complex in its output. Simple enough to understand in one sense, but complicated enough to not be understandable in another. Because there’s a handy database of a lot of good tunes, in a form that’s easy for computers to understand. Because (perhaps – http://www.comp.dit.ie/bduggan/) the author has a personal interest in ITM.
As a non-expert, I think that I’ve heard that this (music generation) is a current research area in AI. I know that Douglas Hofstadter is/was doing the same kind of thing with early classical era music, by trying to create music that can be mistaken for Mozart. The point isn’t to fool people into thinking they’re listening to Mozart, but instead to learn more about what it is about the music of Mozart that speaks to us the way that it does, and to understand more about how our minds and computer minds work.
I’m pretty sure (but too lazy to look up) that was just someone else’s work he was reporting on. It doesn’t sound anything like his usual fields of AI research that I’ve read about. (Editted to add – just searched through a couple of his books, and I only see him talking about other people’s research – and I get the feeling that he would consider the sort of thing we’re talking about here to be very far from the sort of AI he is interested in.)
But yeah, if I were doing computer science research today, this is exactly the sort of thing I’d love to be doing. Except I don’t know about the “case based reasoning” part – if that was around when I was studying AI 15 years ago, I’ve completely forgotten about it.
And if it works at all, the results can hardly be worse than some of the crappy modern-composed tunes I’ve heard out there. 
From the author’s web page:
I am a lecturer in the school of computing at the DIT in Kevin St. I hold a first class honours degree in computer science and software engineering from the University of Dublin (studied at the DIT) and a masters degree in Information Technology for Strategic Management (DIT). My masters was entitled “Strategies for Enterprise Voice Enabled Web Systems” and I have published several papers in the area. I am presently working on a PhD. My current working title is “Modelling Creativity in Traditional Irish Flute Playing”. I am developing a system that can articulate traditional tunes in the style of various well known musicians. It will be a tool to help beginner musicians learn the ornamentation and variation techniques of master flute players.
His page also features a photo of him playing the flute. I think this research paper was born out of an attempt to combine business with pleasure. His attempt to mimic traditional ornamentation usage probably has a greater chance of success than the composition project but it will be interesting to see what the results of both studies are.
The phrase may have been Kelly’s but it’s use to criticise the playing of a fellow musician was all MacMahon’s. I dismiss it for it’s intention rather than for the method of it’s delivery.
Thanks Peter for the other quotes from John Kelly, they are priceless.
On another tack, have we discussed John MacDougall, the Cape Breton fiddler who’s composed ca. 33,000 tunes? Against his will, too. Might be a form of psychosis. Psychosis takes many forms.
Knowing the author, I can assure ye that he has no misconceptions of a computer being able to create what could actually be called music.
I don’t know I think he had a point but I hardly know Vallely’s music. McMahon has strong opinions. A while ago he played with the Kronos quartet. In the television documentary made about that one of the audience uinterviewed after a concert remarked: ‘if he ever goes off again against some other musician, we can tell him ‘we saw you tear the arse of Port na bPucai’ with the Kronos quartet’'.
He’s both on the tellie and the radio tonight, if you can get RTE.