Thursday, June 11, 2009

How not to research about music

Wolfram Alpha has been big in the news as a possible Google killer. The idea is that specific questions are answered, with an emphasis on mathematical calculations. However, the music section leaves much to be desired. Where are the matrix makers, or PC set calculators? The theory is so basic, it barely gets to the level of the third week of a music fundamentals class (the addition of non-Western scales is pretty cool). This is an example of the lack of respect or understanding many scientists have about the complexity of music theory. I'm reading Fred Lerdahl's Tonal Pitch Space right now, which is full of formulas that could be incorporated into a true Wolfram Alpha music section. Or even giving all of the possible neo-Riemannian transformations that connect two chords would be great. The section on the frequency of pitches would be interesting if it included differences based on tuning systems. But calculating interval sizes by rudimentary means? Boring.

2 comments:

Steven Hall said...

Unfortunately our sound-bite news media isn't very thorough when it comes to fact checking. They hear that Wolfram Alpha (WA) is for searching and they know that Google is also for searching so they reason that WA must have been developed as a Google killer.

WA differs from Google and its like and from Wikipedia in that it returns known information about a topic. Google merely returns links to all web pages that contains the word, words or phrase you enter and Wikipedia returns a traditional narrative about a particular topic.

That WA is deficient in its knoweledge of musical matters is more due to its infancy (Google has been publicly accesible over 10 years---WA less than 10 weeks), the developer being a physicist (his circle of contributors would gravitate towards that realm) and other such factors. The WA project is continually adding new data stores so as it matures we can see an increase in knowledge.

There is a "Contribute structured data" link which says "Do you have data, research results, or other information you want the world to know about? It could be a collection of related facts, a table of measurements, a list of formulas, or a set of relationships. In short, it could be any organized body of knowledge that hangs together in some meaningful way."

So, the useful formulae from Fred Lerdahl's Tonal Pitch Space could indeed be added to WA.

Far from boring I think that WA knowing that the notes: C, Db, E, F, G, A, B constitutes a "C Bhairubahar Thaat scale" is pretty impressive. Try find that is Google. Unless someone on a web page states "Oh by the way, the notes C, Db, E, F, G, A, B form a C Bhairubahar Thaat scale" you will be wading through a whacking-girt amount of links.

WA can be combined with Google and Wikipedia wonderfully. Imagine the following scenario. You have a list of notes (such as above) that you ask WA about. You say "wow!" I've never heard of a Bhairubahar Thaat scale so you ask WA about it and it shows what it knows. Now your interest really is piqued so from the WA page you choose the "search the web for BT" link to find out the history, etc. about it.

The question I suppose is can the music community be mustered to help increase the data store of WA so that we can change the title of your blog post to "How to research about music" and we'll be able to find "all of the possible neo-Riemannian transformations that connect two chords". That would indeed be great.

Give WA time to mature and develop. iPods are ubiquitous now but it took years from their first production for them to get that way. I suspect in a few years time we'll be asking ourselves how we ever managed with the primitive web searches of the past.

As I write this I'm saddened that it will be relagated to being a comment about a post. Too bad it couldn't be a guest post as a response much like what used to happen in quality newspapers.

Thanks for you blog posts they apparantly make me think more critically about musical perceptions.

Steven Hall

Scott said...

Steven, thank you for commenting. Your excuse about WA not being a competitor of Google is somewhat true and somewhat false at the same time. While I grant that they go about things in very different ways, both sources are attempting to return information to a user. So they are in competition to be an information gathering tool. And I can see the potential benefits of the WA approach, but it has clear limitations, based on its newness and on the focus of its creators.

What I find frustrating is that clearly no one at WA thought to consult with a music academic. As I said in the post, other than the information about scales, everything else is of such a rudimentary level as to be the equivalent of elementary school math. And this lack of consultation is a signal of a lack of respect for music research. Surely the creators wouldn't put up information about chemical compounds without consulting with chemists and related researchers. I can see by looking up hydroxyzine that it gives some very useful information for a medical researcher, biochemist, or the like. Whereas by inputting the pitches for a whole tone scale, I would want the interval class vector, the PC set class, the Messiaen Mode of Limited Transposition, etc. Instead I'm given the number of semitones between each pitch, the Whole tone Scale name, and two scale names that are just plain incorrect: "C Wholetone scale with Leading tone" is an oxymoron, whole tone scales are identified by their lack of leading tones; "Db Neapolitan Major Scale" is a jazz scale, following the pattern 122221, not like the 22222 scale I gave. It would have the notes Db D E F# G# A# C, so it is indeed close to the scale I gave, but with Db added which is a big deal. This shows the work of someone who thinks they know about music, but knows just enough to make some really bad mistakes. Perhaps WA will be useful for music some day, but for right now it really isn't, except to identify some scales, with questionable answers.