2003-02-14 13:32 (UTC)

WAV is a Microsoft format

It follows from this that such details of the file format as reverse-engineering has established compare unfavourably in rigour, coherence and good engineering practice with papering the floor of a cave full of incontinent bats; leaving it over night; telling a bunch of hippies it's the score for an interpretive dance to be performed on a boat and having a bunch of autistic speedfreaks infer a "requirements document" by watching this from the shore.

But they do seem to have given them binoculars this time, so the quality audit must have had some effect.

Don't believe me?

Linux being Linux, trying to find a tool that plays these things correctly regardless of the endiannesses of the data and soundcard, and the sample width size and frequence and the stereo or mononess of the data is also "interesting", but they didn't start it.

I was writing code to play (with) .wavs just to Find Stuff Out; now I have to write code to use them at all, except I'm not because I'm busy with Real Work today. (The iterative solution of integral equations is more fun to implement than describe, but you probably knew that.)

[Permalink]

2003-02-14 09:13 (UTC)

[Seen on Usenet] Norway Extend Its Thrid Tentacle to Nepal

I didn't read the post, of course, and I know it's a typo, but even so.

You will regret this, Nepal! Svenssen, unleash the Thrid!
A��eee!

[Permalink]

2003-02-13 17:11 (UTC)

Every time

Last night I invented "Phenomenological Phonology", and only partly because it's pretty much impossible to say. (Go on: try it five times fast.)

Google says I'm too late.

[Permalink]

2003-02-13 10:10 (UTC)

Kookbook: Two short planks, lightly toasted.

There's a "refutation" of Chomsky that's popular with certain kooks which assumes that his notion of "deep structure" is a complete semantic encoding of an utterance which could be used to mechanise translation, for example. This is so obviously a ridiculous straw man that I had always assumed they'd made it up for themselves, which shows how much I know:

If we accept the hypothesis that all languages exhibit basically the same deep structure, then we have theoretical grounds for supposing that mechanical translation should be possible using this kind of analysis. In other words, given a grammar of the input language and an algorithm for determining the deep structure of sentences in the language, we should be able to use this deep structure, together with the transformations in a grammar of the output language, in order to generate the translated sentences.
M F Bott, "Computational Linguistics" in New Horizons in Linguistics, John Lyons (Ed.), Pelican, 1970

To his credit, Bott goes on to admit that exactly none of the components of such a scheme are either available or immanent, and he also doesn't claim that Chomsky would endorse such a position. In any case, the suggestion appeared in a respectable source, so I feel entitled to take it seriously enough to disagree with. To see why this it's so silly, let's return to Sapir's Language, which I've linked before and will probably link again because I love it dearly. In his extended riffing on various translations of the sentence "The farmer kills the duckling.", we get this:

In other words, to paraphrase awkwardly certain latent "demonstrative" ideas, does this farmer (invisible to us, but standing behind a door not far away from us, you being seated yonder well out of reach) kill that duckling (which belongs to you)? or does that farmer (who lives in your neighbourhood and whom we see over there) kill that duckling (that belongs to him)? This type of demonstrative elaboration is foreign to our way of thinking, but it would seem very natural, indeed unavoidable, to a Kwakiutl Indian.
[p. 93]

That different languages encode different information as a matter of grammatical necessity is surely obvious to anyone who has ever studied more than one - the stuff that's "left out" from the point of view of the speaker of another language has to be inferred from context.

In case you think this means I've gone over to the side of the tie-dyed tree-huggers and will no longer be slagging off one B. Whorf, think again. But before I explain why, here's Sapir again, on the fairly modest set of tenses provided by English:

Hence "the square root of 4 is 2," precisely as "my uncle is here now." There are many "primitive" languages that are more philosophical and distinguish between a true "present" and a "customary" or "general" tense.
[p. 99]

Now, I like to think I'm as aware as the speakers of any "primitive" language that the square root of 4 is only fortuitously and provisionally 2, whilst our uncles are, in a sense, always with us, but unless you assume a "deep structure" that somehow encodes this information even though English makes no use of it at the level of "surface structure" (i.e., what English-speaking people actually say) then Bott's hypothesis, as I shall be calling it from now on, is going nowhere.

But, crucially, I do understand the distinction between statements that are true in general and those that are true (only) at present, even though I do not distinguish between them syntactically. It seems to me that there's an interesting parallel between Bott's flawed hypothesis and the Whorf's dafter assertions of linguistic determinism: they both assume that individual statements encode the whole of the speaker's understanding of that which is spoken about.

This is such a silly idea that I can only assume they are both implicitly indebted to the taboo against "Mentalism", which was originally a pejorative term for any theory based on the assumption that people had mental states. American Structuralist linguistics under Bloomfield had rejected "Mentalism" as unscientific, in that other people's mental states are not susceptible to empirical deduction, and was allied with a behaviourist psychology that denied the existence of mental states at all. Currently, I am unable to see the point of any of that, but if you did want to avoid mental states in language then you would presumably want each statement to be as self-contained as possible, and this certainly fits well with Whorf's and Bott's hypotheses.

Personally, I have no problem with mental states - I self-identify as a phenomenologist, even - and I see no reason to expect the semantic inventory of different languages to coincide, any more than their phonemic inventories do. Maybe different languages cultivate different habits of mind to the same extent as they do different habits of mouth, I could go along with that. (I'll be calling this Structural Semantics, though, and not "Whorfian" anything.)

While there isn't much of anything in the phonemic inventory of any known language that I can't do with my mouth - I spend long winter evenings with phonetics textbooks, don't forget - it's another question altogether to internalise the mouth-habits of Swedish (or even French, and after all these years) to the extent of achieving "unaccented" speech at conversational speed. And while I would expect that there isn't much of anything in the semantic inventory of any language that I couldn't learn to do with my mind - with no evidence at all, of course - it would be another question altogether to internalise the semantic habits of, say, Kwakiutl to the extent of not making grammatical slips at conversational speed.

It's one thing to "know" facts about a language at an intellectual level; it's a lot more trouble to "internalise" them so that they become part of the everyday furniture of your mind. I can still get flustered enough to lose the ability to put definite articles at the end of Swedish nouns, but I have yet to propose a distinctively Scandewegian - and deeply exotic - sense of definiteness.

[Permalink]