Language Translation: Human -vs- Computer

Written by:  • Edited by: Rebecca Scudder
Published Nov 30, 2008
• Related Guides: Indirect Objects

A comparison between machine translations and human-direct translation.

Translation: Human-Machine Comparison

If you've ever used Babelfish ©, the online translator, then you know that you can get a pretty good sense of the GIST of a text block IF there are no idioms in it and IF you can choose the correct FROM-TO language sets. Choosing the correct language sets, of course, has a direct corollary in the human world, so there is little to discuss there.

It's noteworthy, however, that the kind of machine translation available at Babel Fish, World Lingo or Free Online Translator still has stringent limitations imposed on the translation effort because of considerations of speed (over the Internet, speed of translation once the words and phrases get there, and speed back to you) and because of the attempt to render a meaningful translation to you at all!

Because of this, most translators online perform in a modified word-for-word form, with a few syntax rules applied and a few hundred of the most common idioms programmed into the system Nevertheless, the importance of context is difficult to overestimate. Negative satirical sentences add up to a massive thumbs down for most readers, but mechanical translators usually translate them in their obvious, thumbs-up meaning.

Here's another example of the extent of the difficulty of online or machine translation: seven words can be arranged in at least 343 ways, even though many of those will not make obvious sense in a meaningful manner. Yet that jump from 7 to 343 (7 x 7 x 7) shows some of the large numbers of possible meaningful combinations which need a pruning algorithm after all those branches. Don't misunderstand, human intelligence is behind the 'machine' translation, inherent in the software parser and meaning-derivation schemes programmed INTO the software, but the amount of brute computing strength needed to parse a complex sentence, with negatives and adjectival clauses and adverbial phrases and indirect objects or indirect quotation... can be daunting even for a Cray supercomputer.

Hence the prevalence today of word-for-word+more that's in wide use online. It gets the gist of the selected text, or website or quotation even if it reads in a lumpy, foreign, brain-fried sort of way conducive to much rib-clutching laughter.

It's worse when idioms are thrown into the mix. Imagine, the sentence reads technical and straight-forward, and then hits an idiom, like this: "The 3.2-liter engine will have 8 cylinders, dual overhead cams and a manganese-epoxy engine block... are you pulling my leg?" The parts of an engine can be handled in most modern languages that have vocabulary for modern technology, but the 'pulling my leg' idiom makes NO SENSE AT ALL, translated word for word into either French, Korean or Thai... I know, I've done it and then had to struggle with blank-faced, quizzical looks which showed my listeners were really questioning my sanity.

So some online translation is possible today, and machine-translation is improving almost daily, due to the professional efforts of teams of interpreters, programmers, linguists and translators. But the depth of human experience that a mature (or even young adult) interpreter can bring to the whole mix is in no immediate danger of being made redundant or obsolete any time soon!


 
blog comments powered by Disqus
Email to a friend