Friday, March 13, 2009

Universal Translators

I’m delighted to introduce a guest blog by linguist and SF writer Suzette Haden Elgin. Among many other works, she’s the author of the Native Tongue trilogy (in which she invented a language designed for talking about the life experiences of women), the Coyote Jones novels, the Planet Ozark trilogy, and the nonfiction series of Gentle Art of Verbal Self-Defense books. THE COMMUNIPATHS, the first Coyote Jones novel, can be read free online at An updated edition of THE GENTLE ART OF VERBAL SELF-DEFENSE, incorporating references to cyberspace communication, is going to be published by Barnes and Noble on March 31. Her website is You can find the link to her blog on the sidebar of this blog. Here she’s posting on the problem of the Universal Translator:

The "Universal Translator" has been the magic wand of science fiction for a long time; reader tolerance for any discussion of language differences was until very recently so limited that you almost had to have a UT, just to let you get on with the story. The problem for linguists has been that UTs were unquestionably part of the fiction -- a "Shazam!" device that writers and readers agreed was needed -- but they were presented as if they were part of the science. Nobody bothered to include any sort of disclaimer; nobody had their characters talk about how inadequate the UT was, or how inadequate its translations were. [If you know any examples of that being done, I'd be very grateful if you'd share them with me, by the way.] And the progress that's been made in machine translation in the past decade has only made things worse. The stories you read about the handheld translation devices now being used by the military, for example, would lead you to believe that the Star-Trek-style UT now exists.

Maybe a couple of examples -- a very simple one and a more complicated one, both restricted to human languages -- would be worth posting. Translation means taking a sequence of Language X and providing a sequence of Language Y that is what a speaker of Language Y would say/write/sign in the same situation. So, in a situation where an English-speaking parent would tell a child, "Be good!”, a French-speaking parent would instead say "Be wise!" That doesn't mean that the word for "good" in French is "wise," nor does it mean that one or the other of those utterances is "only an idiom". It doesn't mean that the French concept of wisdom is equivalent to the English concept of goodness. It just means that "Be wise" is what a French speaker would say/write/sign in a situation where an English speaker would say "Be good." That's the simple example. And we now have the technology necessary to construct a handheld translation device that would contain an assortment of ParentSpeak Utterances (like "Be good!" and "Behave yourself!" and "Go to sleep!" and "Go to your room!" and "Don't eat that!" and -- I hope -- "I love you with all my heart") in a whole array of different languages. That sort of machine translation is surprisingly good when the languages involved are closely related, or when the semantic territory is carefully restricted, or both. It's like the job I had as a translator in the 1960s, where I translated a law firm's correspondence in French, Spanish, German, and Italian into English and wrote the responses. It wasn't that I knew all those languages well; it was just that 99 percent of what the letters said was predictable in advance from the context, and I had a book (primitive handheld device) with the necessary equivalents in all five languages.

And then there's the more complicated example. Suppose the question in a dialogue is "What were you doing all morning?" and the English answer is "I was riding a horse." The utterance that a Navajo speaker would use in that situation is, literally, "A horse was animaling-about with me." The utterance a Hopi speaker would use is, literally, "I was using a horse to move about with." (And of course the utterance a French speaker would use is "I was on a horse.") We have the technology to construct a handheld machine translation device that would provide an array of utterances about riding horses in all four languages and a batch of others. But would that let us carry on multilingual conversations about horseback riding?

1 comment:

  1. Elgin's novels and non-fiction have always been at the top of my recommended list. I rave about her work with language, especially on the cutting edge of SF worldbuilding.

    If you've read any of my posts here oon worldbuilding, go get everything by Elgin that you can lay your hands on.

    It's part of the core material you need to build a galactic civilization -- or even a world-girdling Unified Earth.

    In all my SF, I try to deal with the language issue in various ways that facilitate the story and don't misdirect attention away from the theme. But if I use any kind of translating device, or even a person who knows both languages, I generally inject a plot problem derived from the difficulty of "translating" anything as subtle as language.

    I remember on Andre Norton story in an anthology, but not the title. It was about a young steward on a space-bus passenger ship full of civilians. There was an accident. His job was as a Herald to speak to the paniced passengers in their own languages.

    I have always wanted to write a story about an Interstellar Herald. To a certain extent, that is a factor in my novels MOLT BROTHER and CITY OF A MILLION LEGENDS.

    And of course, Universal Translator malfunctions figure into various episodes of my Kraith stories (Star Trek alternate universe fanfic posted at )

    The generally accepted theory of Star Trek's Universal Translator is that it worked on brain waves, what you might consider telepathy, and was thus capable of decoding and learning languages it had never encountered before.

    Spock's human mother is generally credited with inventing it.

    If I were building a "world" based on the Universal Translator, I would make it originate with some truly alien, perhaps long dead, civilization and be a puzzle to the modern interstellar sciences.

    But it would ALWAYS be a plot element because it wouldn't be 100%reliable.

    Jacqueline Lichtenberg