as Google Translate matures
I took the following three posts from newspapers in Mexico, France and Korea and crunched them through Google Translate, when I heard, from a student, that Google Translate was getting better. It would only make sense that Google Translate's skill at making a difficult language comprehensible to us, English speakers, would increase, as is their skill, certainly, at going the other way, or from going from one of these languages to any of the others.
Before I bring up some questions about what is happening, I'll point out the obvious. We are looking at an equation that looks like X > 1 (X approaches 1) where its rendition gets steadily closer to standard English as we know it, given enough time and attention. Presumably GT has put more time and attention into translating Spanish into English than, say, Korean into English. Presumably it has put even less attention into the more obscure languages, i.e. Estonian, and their translation into, say, Korean or Spanish. Having to translate through another language (say, Estonian -> English -> Korean) would be an inconvenience which would more than double the translation glitches.
But my point is: We are looking at a moving target, as they are getting better all the time, not worse. If they just give you the straight translated words (as the Korean post does, pretty much) that's still better than nothing. If they rearrange some sentences grammatically, that's already an improvement. But they are getting better, not worse. Second, presumably every student knows about this service. Any chunk of language can be translated completely and instantly, any time. We can do pretty well cruising the net and simply GT'ing every single Spanish-language article, as we get pretty lucid English in response; it's passable, clear, almost usable in a formal English setting. The Korean one, not so much. It's pretty clearly butchered English, and we have to wade through it.
translated from Korean (Chosun Ilbo)
translated from French (France24.com)
translated from Spanish (La Jornada)
Now here are my questions: First, as a practical matter, for most businessmen, it doesn't have to be perfect. As a document reaches a certain point, and I would say that the GT version of both French and Spanish seem to be at that point, you say, ok, that's good enough, I don't need a translator any more, except to write contracts. GT is, as my student said, "pretty good these days." So to some degree the entire need to learn has been completely undermined. The ESL market - gone. The need to actually know what something says - why bother?
Second, you'd obviously have these translated documents that make the author look reasonably fluent, when in fact the author can't say a word about what he/she has written. One question, and a "false fluency" would be revealed. We could define faux fluency as the appearance of being fluent, completely or reasonably fluent, when in fact the author has no real fluency (it might be pointed out that this is possible in the real language-learning world as well, where I once practiced a fluent phrase of French, got the pronunciation pretty well, and received in turn a torrent of incomprehensible French from a speaker who presumed I was fluent). The dangers of faux fluency are obvious; people assume you are fluent, and sometimes you don't even hear that they say that.
Finally, I remain curious about the learners who are aware of the technology from the very start; who use it, can't avoid it, and try to learn a language by overpowering the force of technology, or by watching the machine and doing what it does, or by in other ways adjusting to the immutable presence of a machine's actions on language. What effect does it have? Are these learners better off canning the machine?
An industry has been built upon crunching songs through GT and back, and making semi-poetic representations of English that make sense in a kind of truncated, dismembered way. This is the grammar-soup world that all intermediate-level, non-fluent word translators live in and have to adjust to.
Before I bring up some questions about what is happening, I'll point out the obvious. We are looking at an equation that looks like X > 1 (X approaches 1) where its rendition gets steadily closer to standard English as we know it, given enough time and attention. Presumably GT has put more time and attention into translating Spanish into English than, say, Korean into English. Presumably it has put even less attention into the more obscure languages, i.e. Estonian, and their translation into, say, Korean or Spanish. Having to translate through another language (say, Estonian -> English -> Korean) would be an inconvenience which would more than double the translation glitches.
But my point is: We are looking at a moving target, as they are getting better all the time, not worse. If they just give you the straight translated words (as the Korean post does, pretty much) that's still better than nothing. If they rearrange some sentences grammatically, that's already an improvement. But they are getting better, not worse. Second, presumably every student knows about this service. Any chunk of language can be translated completely and instantly, any time. We can do pretty well cruising the net and simply GT'ing every single Spanish-language article, as we get pretty lucid English in response; it's passable, clear, almost usable in a formal English setting. The Korean one, not so much. It's pretty clearly butchered English, and we have to wade through it.
translated from Korean (Chosun Ilbo)
translated from French (France24.com)
translated from Spanish (La Jornada)
Now here are my questions: First, as a practical matter, for most businessmen, it doesn't have to be perfect. As a document reaches a certain point, and I would say that the GT version of both French and Spanish seem to be at that point, you say, ok, that's good enough, I don't need a translator any more, except to write contracts. GT is, as my student said, "pretty good these days." So to some degree the entire need to learn has been completely undermined. The ESL market - gone. The need to actually know what something says - why bother?
Second, you'd obviously have these translated documents that make the author look reasonably fluent, when in fact the author can't say a word about what he/she has written. One question, and a "false fluency" would be revealed. We could define faux fluency as the appearance of being fluent, completely or reasonably fluent, when in fact the author has no real fluency (it might be pointed out that this is possible in the real language-learning world as well, where I once practiced a fluent phrase of French, got the pronunciation pretty well, and received in turn a torrent of incomprehensible French from a speaker who presumed I was fluent). The dangers of faux fluency are obvious; people assume you are fluent, and sometimes you don't even hear that they say that.
Finally, I remain curious about the learners who are aware of the technology from the very start; who use it, can't avoid it, and try to learn a language by overpowering the force of technology, or by watching the machine and doing what it does, or by in other ways adjusting to the immutable presence of a machine's actions on language. What effect does it have? Are these learners better off canning the machine?
An industry has been built upon crunching songs through GT and back, and making semi-poetic representations of English that make sense in a kind of truncated, dismembered way. This is the grammar-soup world that all intermediate-level, non-fluent word translators live in and have to adjust to.
posted by tom at 11:21 AM
0 Comments:
Post a Comment
<< Home