There is no such thing as machine translation

In a recent Maclean's article, the possibility of having a "universal translator", that is, a machine or computer for translating both written and spoken language, is explored in depth.

While some sources interviewed for the article are quite optimistic, one has to keep in mind that those are computer experts who don't know the first thing about the process of "human translation" – or language as such, for that matter.

Fortunately, the article also quotes someone more knowledgeable:

Denis Bousquet, vice-president of the Canadian Translators, Terminologists and Interpreters Council, says the technology would require artificial intelligence in order to do the job, and that is still generations away. "Machines are a bunch of zeros and ones and language is way more interpretive than that," he says. "Emotion and colloquial expressions come into play and you can't translate those with a strictly mechanical process."

Exactly.

The problem with automated translation is that language is one of the most human activities imaginable and therefore cannot be captured by zeros and ones. While individual words can be translated by a machine, or even very short and standardized phrases ("How are you?"), there is no way a computer can understand context. This is why human translators have to read the whole text to be translated first, understand and internalize it, and then translate it. Since the functioning of the human brain is not understood yet as to how language is processed by the brain in terms of words, syntax and, more importantly, nuance, there is no way of programming machines to imitate the very processes that scientists do not even understand yet. It will be several hundred years, if ever, before machines can fully replace human translators and/or do away with the need to learn foreign languages.

Comments

[Judy Jenner]

Agreed 100%. Just like a computer can't write its own code (at least that I know of), a computer can't reliably evaluate syntax, tone, etc., even though it might be able to do basic conjugations. Humans are not replaceable, at least not in translation.

[Werner Patels]

I don't doubt that there will be many areas where humans are replaced by machines and/or automated processes in the future, but translation and language will not be one of them.

[Denis L. Bousquet]

Dear Werner: That you qualified me as knowledgeable is quite an honour, for which I thank you. Your blog members might be happy to know that at the last ATA conference, held in early November, I had a conversation with the IMTA (International Machine Translation Association) president during which she expressed that "we want to ask translators what they look for in a TM (translation memory) or MT (machine translation) tool." The research done is always based on their clients requirements and not on the end-user needs. So this is a nice change in point of view. CTTIC will be collaborating with IMTA in putting together a short survey to compile information in that respect. Check our website in the early months of the new year as we will publish information in that respect. In closing, I want to congratulate you on your very informative blog. Keep up the good work.
Best regards.
Denis L. Bousquet
CTTIC President

[Werner Patels]

Thank you, Denis, for stopping by and sharing the valuable information about your survey.

[Emily]

Human language is a fine craft involving very fine brain processing, and no machine can copy it. I really don't understand why it is so hard for some people to understand that language is not some strict math formula. I mean, I don't mind if they want to teach their kids to speak and sound like robots, but my kids will acquire natural language, which is a very complex and refine brain process happening naturally by itself. Even the attempt of this is silly, it is like creating a machine that will teach our limbs to grow.

[Mike Unwalla]

I agree with your sentiments. Universal machine translation is not going to happen for many years, if ever. However, many businesses use machine translation successfully.

Werner wrote, "The problem with automated translation is that language is one of the most human activities imaginable and therefore cannot be captured by zeros and ones."

Emily wrote, "I really don't understand why it is so hard for some people to understand that language is not some strict math formula."

Language can be formalised. Words can be given restricted and specified meanings. Sentence structure can be restricted. Ambiguity can be removed. In such cases, machine translation gives excellent results.

Even with less restricted cases, words can be used with their primary meanings, and language can be simplified. In such cases, machine translation gives good results. For more information, see www.international-english.co.uk.

I think that translators would have more credibility if they acknowledged that in some cases, machine translation gives sufficiently good results. Machine translation may never give good results with poetry or literature, but that is no reason to reject machine translation.

[Werner Patels, M.A.]

Yes, in some basic ways machine translation can produce results, which, however, must still be edited by a human translator. Because even texts with very basic syntax and/or words used in their primary meanings contain nuances and other elements ("reading between the lines") that require a pair of human eyes and a human brain (and life experience).