“I know that you believe you understand what you think I said, but I'm not sure you realize that what you heard is not what I meant.”
--Attributed to Robert McCloskey, U.S. State Department spokesman, by Marvin Kalb, CBS reporter, in TV Guide, 31 March 1984, citing an unspecified press briefing during the Vietnam War.
“Counting words is merely a device and while quick, easy to grasp and convenient, a poor one…Turning texts into segments, calculating repetitions, quantifying the value of those segments based on so-called repetition, has directly contributed to what I believe has been referred to here as ‘commoditization’.”
MEANING IS NOT A COMMODITY
Previous articles on this blog have addressed the issue of “selling words” to refer to the way most freelance translators price their work. While this concept has always been flawed, it represented a way of negotiating prices with purchasers of translation services which, with the advent of CAT tools, became even more problematic because translators found themselves faced with demands for discounts on previously translated words or “repetitions” and words that were “close to” words that had been previously translated or what came to be known as “fuzzy matches”.
All to no avail, I have tried to determine the targeted user of these “CAT tools”, a term which is used for software programs consisting, in their simplest form, of a translation memory and a terminology database. My gut feeling, based on what I have gleaned from those who are pushing their use, is that these programs were designed to help the individual translator, who would develop an ongoing “translation memory” consisting of matched source- and target-language segments. This translation memory would work in conjunction with a terminology database which the translator would constantly update. The translator was led to believe that he/she would “never have to translate the same sentence again”.
For many translators, it was a rude awakening to discover that translation agencies were developing translation memories for all their clients and that they, the translators, were expected to give discounts for previously translated words, whether or not they agreed that those previously translated words conveyed the meaning of the source document. Sadly, most translators simply complied. After all, there was no financial incentive to make changes and the only way to survive financially was to produce more words albeit at a reduced rate.
Texts were turned into segments that were matched with previously translated segments, previously translated source-language words were matched with target-language words to arrive at a “translation”, which supposedly conveyed the meaning of the source-language text. The translator became a “language engineer” who manipulated segments and marveled at the way the software could reproduce formatting and put all those segments back together in a way that seemed to produce a target-language document that was identical to the source-language document.
Machine translation, which incorporates some of the same technology, also works with words. This technology may be based on rules of grammar and dictionaries or it may be statistically based on natural language usage, but ultimately it is based on words which have denotative and connotative values determined by context.
All well and good were it not for the fact that writers in the real world use language that is not only idiomatic and/or idiosyncratic, but rather is language that I have come to call Humpty Dumpty language, a language characterized by the concept “a word means exactly what I choose it to mean”. So, what does the human translator do when he/she encounters a word in the source language for which any of the possible alternatives in the target language would be somewhat bizarre? I tend to think that the human translator would opt for the intended meaning. No automated language program is capable of making that determination, and it is not something that can become part of a translation memory or a terminology database.
In reality, what seems to have gotten lost is that the role of the translator is to communicate meaning. While that meaning is of necessity couched in words, there is no direct correlation between the words in one language and the words in another language, nor can we be sure that the words used by a writer in one language convey the meaning that he/she intended to convey. We are all aware that it is becoming increasingly difficult to find a document that is well written in any language.
So, in the end, meaning is not a commodity. It cannot be reduced to words to be sold to the highest (or lowest) bidder. Although meaning is expressed in words, those words are constantly changing. Any attempt to automate the expression of meaning is bound to fall short.
The automated language industry is actively trying to involve human translators in the machine translation process. In reality, they need the work of human translators to provide the matched segments that is the basis of statistical machine translation. However, they seem to ignore much of the advice that comes from translators. In discussing the role of translators, Fred Hollowood of Symantec Corporation said: “I relied heavily on the quality assessments of translators. They were not always favorable. It was some time before I learned to temper these evaluations with automatic metrics and user evaluation of MT output.”
Automatic metrics will be the subject of a future article, but user evaluation of MT output is something that can be compared with user evaluation of human translation. Oftentimes it is not the end-user of the translation who evaluates it but the entity contracting the translation. This entity is not always capable of determining how it will be understood by the end user.
There is no doubt that the translation industry is in a profound state of change. We cannot cling to the past, neither can we ignore it. The merchant translator that Bernie Bierman mentioned in previous articles is gone forever. The freelance translator has no “knight in shining armor” to protect his/her interests. But we do know that our profession is not and should not be based on the sale of words. The role of the translator is, has always been, and hopefully will always be the communication of meaning and that is something that defies commodification.
I personally believe that CAT tools (Trados, etc.) as we know them, like the 8-track, will soon be part of the past. This is not true of machine translation, and it is a force that all human translators must reckon with. Dealing with the automated language sector (their term) is not easy in that they have their own “language”. Navigating their websites and discussion groups involves constant googling to determine the meaning of acronyms and arcane language. Nevertheless, I feel that only by attempting to come to grips with their goals, whether or not they are achievable given the idiosyncratic nature of human language, can we achieve a balance in human communication.
Right now, though, what we need is a real dialogue about post-editing, what it is, how it fits in with traditional translation, and how it should be remunerated. Ultimately, that is the bottom line because if translators are expected to contribute to the machine translation process, they should be compensated for their efforts.