Gustav Brock
Gustav at cactus.dk
Tue Nov 23 03:27:14 CST 2004
Wonderful! Thank you so much Arthur. /gustav >>> artful at rogers.com 23-11-2004 00:18:16 >>> 1. There are a pair of classic sentences that are used frequently to derail machine translators: a) Bill sent me a bill which I forwarded to Bill. b) Buffalo buffalo Buffalo buffalo Buffalo buffalo. (This one I capitalized to help you through it, although it still takes some work. To save you the trouble.... there is an English verb, to buffalo. It means to baffle, or perhaps to bullshit. As in, when the client asked question X, I buffaloed her. A particular variation on said action is known as the Buffalo buffalo. This is a version particular to the city -- as opposed to say the New Orleans buffalo. So, certain buffalo who happen to reside in Buffalo do the Buffalo buffalo to other buffalo who also happen to reside in Buffalo.) There are many other variants on this problem, and to give my source credit they can be found in Steve Pinker's classic book "The Language Instinct". (For those unfamiliar with it and who also happen to wonder what the hell Noam Chomsky was on about, said book contains the most concise explication of Noam's basic theory that I have ever read.) Here are a few more nuggets for your translation-machine: Deer Kill 180,000 Queen Mary Having Bottom Scraped And of course, the old saw, from Marx (Groucho not Karl), "Time flies like an arrow. Fruit flies like a banana." An even more remarkable one, also courtesty Mr. Pinker.... here are 2 sentences, each spoken by a different person (presumably married to each other): a. I'm leaving you. b. Who is he? This one is my favourite. The other ambiguities are resoluble by parsing the sentence in question carefully. But this latter is remarkable in that: a) none of the words assembled convey a larger meaning (as is the case, for example, in "The New York Yankees"). b) all of the words assembled convey no inherent meaning (you have to assume a HUGE amount of cultural baggage in order to interpret this conversation; further, you have to assume certain mores, certain cultural perspectives, certain sexual preferences and so on, in order to make any sense of this snippet of conversation -- i.e. regardless of the gender of A, we can deduce that A is only interested in males, but that may be a faulty deduction; A could be bisexual and B unaware of this; it is also possible that A is leaving B not for someone else but simply because living with B is intolerable; and so on). IOW, virtually NONE of the meat in this conversation is contained either in the individual words or in their assemblage; ALL of the meaning is contained in the minds (and by extension the cultures) of the speakers. I have a lot of European friends, and almost all of them have commented at one time or another on the elasticity of English. Consider the word "fuck". It can be used in so many ways that it boggles the minds of newcomers to the English language. I fucked a lovely woman last night. I fucked a business competitor this morning. I tried to repair my motorcycle but fucked it up. Fuck you! Fuck that! Fuck you and the horse you rode in on. Fuck me. And so on. There's a common (and I might add falacious) maxim that the Inuit have 28 words for snow. This is true only in the sense that we have 100 words for cigarette or automobile (Honda Prelude, Dodge Caravan, etc.) The example cited above is the opposite phenomenon, as exemplified by another old riddle: Why is making love in a canoe like drinking Budweiser? Because it's fucking close to water. Insofar as writing remains utterly ambiguous (a pretty tough proposition, but let's let that pass), then translation-engines have a slim chance of accuracy. To the extent that language is a joy, a thing to be savoured, played with, caressed, teased and fondled, then machine-translation has virtually no chance of success. If we go no further than translating "File | Open..." then we may be safe; ditto for various error messages.... but even there, I have to point out that once I taught a course, one of whose students was confined to a wheelchair, and when a message came up saying something like "Invalid parameter", she said, "I'm an invalid." Which in turn led me to consider what a horrid word that was to describe people in her condition. Such persons are not invalid; they are just as valid as you and me. Yet another example. A close friend and colleague of mine, from Britain, gave me a copy of his new software to test. A couple of days later he asked me what I thought of it. I replied, "Not bad!". In Canada this is a compliment. In Britain this means that I couldn't find anything truly despicable to say about it, but was overall underwhelmed. His look of shock make me realize a cultural gap here of serious proportions. Chevrolet at one point release an automobile called "Nova". In Spanish this means "Does not go." (Neuve is the word for New.) All of which goes to say, do not trust translation-machines. If you need to translate, find a human. One final example, taken from the annals of machine-translation literature. They had an English to Russian translator and its opposite. They fed in an English sentence, translated it to Russian, and then translated it back. a) The spirit is willing but the flesh is weak. b) The vodka is good but the meat is rotten. Arthur