Linguistics: April 2008 Archives

I went into a second-hand bookshop on the way home from uni yesterday. (Pickerings Books, 30 Buccleuch Street (just on the corner with Buccleuch Place). It's closing down, unfortunately.) I usually look for books on Philosophy, Religion, and Language, although I'm open to all really. Sometimes you can get real gems.

Of course, everyone's definition of a gem differs. While my interest in language are mainly theoretical and phonological, I do have a penchant for language descriptions and obscure languages, and the idea of descriptive fieldwork seems very romantic and adventurous to me. (You'd be like Indiana Jones, but a linguist instead of an archaeologist!) So I stumble upon David Watters' A Grammar of Kham, a language spoken by roughly 45,000 people in western Nepal. How much? £5! You can't say no for that price. Especially when it's in such a good condition - hardback, no notable scuffs, paper practically as-new. I get home, and find out that Cambridge University Press, the publishers, expect £96 for the book, and that amazon.co.uk are charging £91.20.

Slightly less impressive, but still a good saving, I got Geoffrey Kimball's Koasati Dictionary (also known as Coushatta, a language spoken by around 400 people in Louisiana) for £4, when amazon.co.uk want £64, and the publishers $85.

Do I need these books? No. Do I want them? Yes. Will I use them, and find them useful? Probably. As I was explaining to my flatmate yesterday, I'm not buying them for any pragmatic value. I doubt I'll be stuck in western Nepal any time soon, or be asked to translate some Koasati myths. I got them because I find this stuff interesting. And because they were bargains!

Kimball, Geoffrey D. Koasati Dictionary. Nebraska: University of Nebraska Press, 1994.
Watters, David E. A Grammar of Kham. Cambridge: Cambridge University Press, 2002.

I spent a good two hours last night playing around with Praat, with some rather odd results. Praat is a phonetic analysis program, packed full of features, that I've been using for over two years now. So far, I've mainly been using it for spectrograms, annotation of files, and making measurements. However, I needed to actually manipulate some sound files, so I started playing around with the settings to see what there is on offer. There's a lot, most of it over my head - multidimensional scaling and combine to ParamCurve, anyone?

I ended up making this WAV file (right click, "save target as" to download), mostly accidentally. It was originally a recording of me saying brewed, but has been manipulated in several ways. A few of the sounds (particularly the distorted time parts) remind me of noises in a lot of breakcore music I listen to (for instance, at around 1:45 this video of Venetian Snare's Szamár Madár (yes, it's meant to sound like that)).

The particular manipulation I wanted to do, was to take a sound file, and take a particular section of that file (from, say, 25ms to 150ms), and change the duration of that section (from, say, 125ms to 100ms). Reading the manual it seemed that this was indeed possible, but to do it with any great precision would require a script. So I figured, let's write this script, and since I'm going to the trouble of writing a script I might as well make it automate most of the process for me. As such, the script will manipulate the duration between the specified time points of the selected sound file, and, optionally, write the output to a .WAV file.

This is the first Praat script I've written from scratch (I've done a few others, mainly edits of other people's work to better suit whatever I'm working on), and if you want to see it, it's here. Hopefully the comments in the code make it fairly transparent, and I think its operation is relatively simple.

Okay, technical stuff now. Look away now! Don't say I didn't warn you! Beware! ... Now that I've scared away the non-geeks, let me talk a little about the extension possibilities of this script. Currently it's fairly limited - it only outputs to WAV, for example. It should be fairly easy to change what it outputs to (or even make a range of options appear on the form). In its current form, the user has to specify the start and the end of the area to be manipulated, by entering numbers. Another (probably easier) way to do this would be to take the information from a textgrid file associated with the sound file. Then all the user has to do is to make a textgrid file, add some points where they want the duration to be changed, and run the script. Coupled with a modification that allows for batch processing of files - modifying several files in one go - this would allow for fairly extensive modification at the click of a button.