E X C L U S I V E I N T E R V I E W
(Part 2)
The interview below was conducted between Calgary, Canada and Valencia, Spain
Here is a link to the first part of this interview: https://bit.ly/2lYSfyS
| Calgary, Canada | Valence, Spain |
Our interviewer, Susan VO is a French Interpreter with 14 years experience as a staff member and freelancer with the United Nations, the Canadian Federal Government and in the private sector. She is an alumna of the the School of Translation and Interpretation at the University of Ottawa, which Brian Harris helped developed. She was Linguist of the Month on this blog: her interview can be found here and here.
Our guest interviewee, Brian HARRIS, has just celebrated his 90th birthday. His long, interesting and prodigious career in the theory and practice of translating and interpreting, as well as his strong interest in history, is reflected in this interview. Special mention should be made of the fact that he coined the term 'translatology' for the scientific study of translation. (In the 1970s, a French professor of translation, René Ladmiral, introduced traductologie in French. Traductologie caught on and was soon borrowed into other Romance languages as traductología, etc.; translatology never caught on and was eclipsed by ‘translation studies’.) Natural translation is Harris' most important contribution to translation studies. In the early 1970s he began to notice that while he was supposedly teaching university students to translate, many people were doing translation successfully without such training; indeed that the untrained translators were doing more translating than the trained ones and often to just as high a standard. Many of the interpreters Harris worked with, including some from the Parliament of Canada had never had formal training. This led Harris to the conclusion that all bilinguals can translate within certain limits. In 1978, he and an assistant, Bianca Sherwood, published "Translation as an Innate Skill", which has been described as the seminal article on natural translation.
Brian lives in Valencia, Spain with his wife and cats. His blog is accessible at UNPROFESSIONAL TRANSLATION
---------------------
Susan Vo: How did the theory of Natural Translation play a role in developing the School of Translators and Interpreters at Ottawa University and how was it received by the academic community at the time?
The latter 50 years of my career have been dominated by missionary work for the Natural Translation Hypothesis (NTH), which is of more lasting importance than all the rest. I call it a hypothesis because there’s as yet no definite proof of it, but the indications are strong.
We can divide it into several propositions. The first is that all bilinguals can translate. I wasn’t the first to assert this; my mentor in translation studies, the Bulgarian semiotician Alexander Ludskanov, wrote it a decade before me. What’s more, he explained the difference between natural (i.e. untrained) translators and professional ones. He said that what we teach in translation schools is not to translate but to do so according to the norms and standards of a culture and a society.
The second proposition is that bilinguals’ universal ability to translate is innate. That’s to say, along with our ability to learn languages, we are born with the ability to translate between them. The key paper on this point is “Translation as an Innate Skill”, which I wrote with my student Bianca Sherwood in 1976 and which is available for everyone to read through my Academia.edu page. The main argument for this assertion is the very young age at which bilingual children start to translate, and to translate quite well; they do it at around three years old and without any instruction from their elders. It’s analogous to the argument that Chomsky uses for innate language competence. We were very lucky, when we started writing the paper, to receive a generous gift of data from an educational psycholinguist in Toronto called Meryl Swain who had been recording a Quebec bilingual boy.
I wasn’t the first either to observe that young children can translate. That distinction belongs to a French linguist named Jules Ronjat who published a study of his own bilingual son in 1913.
But both Ljudskanov’s declaration and Ronjat’s description had gone unnoticed by translation theorists. My contribution was to point out the significance of their work and to continue it.
“Innate Skill” was generally received with scepticism or even outright ridicule by the community of professional translators and translation teachers. On the other hand, it was appreciated by some leading psycholinguists like Wallace Lambert at McGill University in Canada, David Gerver at Stirling University in Scotland, and Kenji Hakuta and his student Marguerite Malakoff at Stanford University in the USA. Also by one influential translation theorist, Gideon Toury , who had a model of his own called Native Translation that fitted in with mine.
Acceptance of the concept has advanced only slowly in the last 40 years, but some aspects of it are now mainstream, or almost. Language brokering studies, which started in the USA in the 90s, opened people’s eyes to the vast amount of translating done by children. The NPIT (non-professional translation) conferences and publications of the last decade have helped brush away the cobweb of misunderstanding in the old saying that “because you are bilingual, it doesn’t mean you can translate” (or interpret, for that matter).All around us, NGOs, manga and computer game publishers, Wikipedia and others depend on crowdsourcing their translations. Of course there’s a tradeoff. Mass production and amateurism can rarely matched skilled craftmanship, but it’s a price to pay to get the translations done.
The third proposition is that there are two pathways –as in other skills—from Natural Translation to Expert or Professional Translation. One is by formal instruction and the other is self-instruction by imitation. The second is the way we learn our first language, and it’s what Toury meant by Native Translation.
Finally I’ve gone back in my blog “Unprofessional Translation”, to an idea that was already held by semioticians like Ludskanov. It’s that what we call translation is the language specialization of a more general conversion of all kinds of signs, and it’s that general ability which we inherit.
Susan Vo: In your own words, with hindsight and observations of current trends, how would you say that Natural Translation and Simultaneous Interpretation are similar? What kind of traits do you believe all simultaneous interpreters inherently possess, (from a cognitive, cultural and even personality standpoint), how do these traits develop, either naturally or with deliberation?
The Natural Translation Hypothesis is a general theory about all translation (spoken, written or signed) and it says nothing that’s specific to simultaneous interpreting or indeed any interpreting. It goes without saying that simultaneous interpreters have to be competent translators, but NTH isn’t concerned with the quality of translations beyond a basic, childlike level; only with whether people can translate. There are too many other factors in expert translation, such as family, schooling, work experience, travel, etcetera. Nevertheless, leaving aside NTH, there may well be features that are natural in the sense that they are, or they develop from, abilities that we interpreters are born with or that develop in us without being taught to us – which doesn’t mean that they can’t be improved by teaching and practice.
The one most commented on is mental speed. In simple terms, simultaneous interpreters have to be quick thinkers, but it’s not so simple. Simultaneous interpreting is not really completely simultaneous. There’s what linguists call the latency, or ear-voice span, typically two or three seconds. But that’s the most simultaneous interpreters can allow themselves if they don’t want to lose part of what the speaker is saying. Not everyone can keep this up. That’s why I and others have insisted on a shadowing test in admission exams. There have been magnetic resonance imaging studies recently that show there may be a physiological factor in mental speed, to do with the coating on the axons in our brains. But that doesn’t prove it’s inherited.
Another one often mentioned is personality. It’s true conference interpreters are performers, because they have to perform live often before an audience of thousands. So I’m inclined to think there’s a connection. Studies of the relationship go back to the 1950s, but without conclusive evidence or proof that it’s innate. So all we can say is maybe.
And the same applies to concentration, split-mindedness, stamina, even ability to work as a team.
As for “current trends”, the hot topic at the moment is automation. It’s true that interpretation only operates at present at the simple level required by NTH but it will improve. And automation is the opposite of natural.
Susan Vo: Machine translation, which had a pivotal moment in 1988, can be said to be the precursor of capabilities being used commonly today and advancing: google translate, translation apps, use of artificial intelligence in linguistic services. What are your thoughts on the role of MT, the role of the human translator, and where we are heading?
My interest in machine translation goes back a long way. It was in 1966 that I was recruited to a team at the Université de Montréal that was doing research on MT for the Canadian National Research Council. We were part of the second generation of MT researchers; the first was in the 1950s. I was recruited as a linguist but I quickly understood that you can’t research MT without some understanding of computers. So I took courses in programming and mathematical linguistics and worked for three years as an assistant to a brilliant French computer scientist named Alain Colmerauer who was later the inventor of an AI programming language called PROLOG. We had some limited success by designing the prototype of an MT program called METE0 that has translated many of the Canadian official weather bulletins between English and French since 1974. But the computers and software of that epoch couldn’t have handled today’s AI. Instead we, like our French and Soviet contemporaries, used grammars and dictionaries.
Then in the late 1980s, long after I’d left MT for other interests and when computers had become vastly more powerful, there was a revolution caused by IBM’s introduction of statistical machine translation (SMT). It became the basis of today’s MT. I had played a small part in its beginnings with some work on the alignment of translations with their source texts, but that work was insignificant compared with IBM’s.
And then in 1996 I was given a new understanding of MT and AI by sheer chance. One of my Ottawa students named Bruce McHaffie came to me with a proposal to explore the use of neural networks for MT. (Neural networks are currently the dominant computer tools for what’s popularly called AI.) I encouraged him and he succeeded in producing a feasibility study for his MA thesis. He was a pioneer; however, he only had primitive neural network software at his disposal and it was more than a decade before networks became mainstream.
As to whether AI produces better results than statistical MT, there is a saying that “The proof of the pudding is in the eating.” Try it for yourself; after all, it’s widely available on the web and it’s free. My own experience is that at present it’s only marginally better. But it does have one major advantage over SMT, which is that it doesn’t need preliminary close alignment of texts. Therefore, over time, there will be much more of it and that in itself should lead to further improvements because AI systems learn by experience.
In the long term, MT still faces problems that current AI cannot solve. One of them was foreseen by the Israeli researcher Yehoshua Bar-Hillel back in the 1960s. It’s the application of non-linguistic knowledge, or what he called encyclopedic knowledge, because we don’t have adequate computer representations of such knowledge. For example, the correct translation of such a simple sentence as “Cross the river” requires a French translator (or MT system) to know whether the addressee is a close acquaintance (Traverse la rivière) or not (Traversez la rivière) and to be sensitive to the difference in usage between European and Canadian French; and also to know whether it’s an ordinary river (rivière) or a large one flowing into the sea (fleuve). Legal translation requires knowledge of legal systems.
But in 1966 we couldn’t foresee MT like today’s, and so we just have to wait for the next 1980s revolution. Anyway, MT has reached a point of no return and the next step is MI (Machine Interpreting). It’s already on the horizon.
Commentaires