Re: [atlarge-discuss] Translation issues

[espresso@e-scape.net]

I realize some of you may be bored by this discussion but I do feel a need to clarify some of the issues again...for the last time, I hope.

At 21:02 -0700 2002/08/18, Jeff Williams wrote:
[me] >> 2) translation memory programs, which in effect match 
>>words or expressions in a document against a database of 
>>previous translations; the results of processing a document 
>>with a TM depend very much on the extent and quality of the 
>>database, and at best require a lot more than "proofreading" 
>>for "small errors" -- more like a systematic line-by-line edit 
>>of the material which, in my experience, takes almost as much 
>>time and effort as translating the whole thing in the first 
>>place.
>
>  Yo are right that this is one such category and that the accuracy of the
>comparison database is if primary interest.  However your conclusion as
>additional personal opinion is not broadly shared.  For instance this
>type of translation software system is what is used by the UN, and works
>very well for them...

TM software can work extremely well in any organization where there are teams of highly-skilled terminologists and translators who prepare the databases and edit the results in parallel with the originals. The problem is that, while the UN and similar organizations understand translation issues well and make sure the process is handled so as to produce accurate, well-written results, most of the organizations using TMs don't. 

Some simply don't understand translation and assume that for every word in their own language, there is one and only one word with an identical meaning in every other langauge, so they expect something like Babelfish or PowerTranslator will just make the automatic substitutions and everything will be fine. This is not a rare problem amongst people who have never learned a second language. It affects a great many businesses trying to sell their products globally, and from my experience of explaining what translators actually do, I can help knowing most unilingual people do make that assumption.

Every now and then, some English-speaking Albertan working for the Canadian government causes a major uproar in the Quebec media by putting one of those "translations" up on a government Web site. (The most recent incident was at Environment Canada but there have been others, alas.) Meanwhile, English-speaking Canadians are offended when French-speaking people do the same. As a Canadian whose federal government is officially bilingual, I'm acutely aware that, while people get mildy annoyed by spelling mistakes or grammatical awkwardness, they get quite fired up when their language is butchered or they can't make sense of the information they need.

>> The exceptions would be things like spec sheets for 47 models of the same type of equipment, formulaic correspondence  containing only grammatically simple sentences, etc.
>
>  Yadda, yadda, yadda...   This is not only not a good comparison it is not
>even in the same category either...  Such analogies/comparisions as a
>argument in response for something you obviously seem to be adverse to
>does not lend itself to being a very strong argument on it's merits or >based on this argument/comparision just above that you have not so kindly >provided...

I'm not sure what you mean here. It's well-known by both translators and makers of translation software that, like any other process, you can save time and money by automating only when the process contains a lot of repetitions. I'm by no means averse to computer-assisted translation: I just don't do that much work involving formulaic expressions and repetitive vocabulary myself: I find little help in having the software translate every instance of "the" into "le" so that I can go back over the results and change the right occurrences to "les", "la" or "l'" as the grammatical context requires, to take a very simple example. 

Though I admit my personal writing style is somewhat complex, I am all in favour of using simple declarative sentences where they will do the job. Back in my youth when I worked in people's offices, I'd have given my eyeteeth for a desktop computer (they took up whole rooms and cost thousands back then) which would spare me from typing the same boring words over and over again. I was thrilled when they invented machines that would let yopu paste in boilerplate text with a couple of keystrokes or knock out a form-letter to dozens of people. But I do care about language -- clarity, accuracy and good usage as well as appropriateness for the destined readership -- and we humans can't say everything we need to communicate using only boilerplate sentences. 

Pick up any reasonably good book (other than something like a technical manual) in any language and read a couple of pages to see how many types of sentences, how many idiomatic expressions, and how many nouns and verbs are used over and over again, as opposed to recurring in different forms or contexts. Those working on development of computer-assisted translation programs are well aware that no existing program fully comprehends human language and how it works. How do you decide whether a particular sentence calls for "say", "says", "said", "was saying"...? It's a very complex intellectual process, even though we do it without conscious thought most of the time, it's not easily reduced to yes/no decisions a computer can handle.

>  I have several years now of using several different types of >software/system based translation facilities.  They have been a 
>watershed of a useful tool in reducing the human impact of 
>translation of a host of different documents both of the legal 
>and intricate variety and of the less specific or intricate as 
>well..

I confess I am not entirely sure why "reducing the human impact of translation" would be a desirable outcome. Cost and time are what most people try to save using these things. But the purpose of translation is to convey the meaning of what is said by a human in one language to another human speaking a different one. Computer-assisted translation programs don't understand either human language. They deal the statistical probability that a word or expression in the source is translatable into a given word or expression in another, with or without other possible translations which are somewhat less probable. 

They're of greatest help in documents where each word or expression has only one possible form and meaning, and is translatable into a term in the other language which is similarly precise in its meaning. The Météo program is a perfect example: temperature, windspeed, etc. are expressed in a fixed vocabulary all meteorologists understand in precise terms in each language and use consistently for all weather at all times. Human intervention isn't needed once the terminology banks and substitution algorithms are set up.

When it comes to something like a legal document or set of bylaws, you simply can't do that so easily. The precise legal meaning of a word as simple as "sale" or "contract" varies by jurisdiction, and it may be further nuanced by the context in which it occurs, sometimes to the point where one needs to footnote a clause with an official translation of the law in jurisdiction A and an explanation of how it differs from jurisdiction B. The words "administrateur" and "directeur" occur often in French bylaws but what is meant by them does not necessarily correspond at all to "administrator" and "director"; in fact, they may well mean "director" and "administrator" instead. Which means you need to know the cultural context in which the source text was written AND the cultural context in which the translation  will be read.

>  No, not necessarily.  Some bylaws are quite simple in the 
>structure and use of language and terms.  ICANN's is one such 
>example, BTW.

I agree with you there, and it's far preferable for bylaws to be written in plain language so everyone can understand them. Alas, that's not always the case. Some I've been asked to translate were written with such a lack of clarity that all I could do was query the client with "were you trying to say X or did you mean Y?" so as to render the right interpretation in the target language.

>  I agree as close to 100% as is possible is what should be the goal 
>and at least a 95% accuracy would be minimal...  Hence using a 
>translation tool to do most of that work getting at least 80% without 
>human intervention, is a huge aid or assistance.

More power to you if you find it so, and presumably one task less on my already-crowded plate.

>> I presume the "TOOLS" you mean are things like Trados, DéjaVu, 
>Star Transit, etc.
>
>  No, those are pure garbage and not worth using...

I'm surprised to hear you say so, since many of my fellow-translators seem to use them very productively on corporate documents or technical projects involving a good deal of repetition. Some claim they save as much as 20-40% of the time required to do decent translations... but, then, that doesn't include the time required to learn to use a program efficiently and to prepare the databases properly so it only makes sense for those who do a lot of the same kind of documents. Meanwhile, the number of translation agencies which require their translators to use one or another of those programs has grown at least 100% over the past two years -- again, usually those who deal in bulk quantities for particular corporations.

>  Bylaws in english is usually easily understood by all native 
>english language readers or speakers.  Hence why in almost every 
>case that I have seen in 23 years of experience, I have not seen 
>more than one version in English, French, Spanish, or Portuguese.

That's not surprising at all. Bylaws are ordinarily written for one organization's use, and international organizations are very much in the minority. 

Organizations which are international in scope usually take the trouble to have their bylaws translated carefully into their official languages so as to avoid regionalisms. The fundamentals of Canadian and Belgian French are no different from the fundamentals in France, and the same is true for the Spanish of Spain and Latin America or the English of the various English-speaking countries. Also, bylaws almost invariably begin with a "definitions" section which spells our precisely how particular words are to be interpreted -- good idea if you want to avoid endless debates on semantics from the more troublesome members. And members of an organization who bother to read their bylaws closely (most don't) always have the possibility of asking the secretariat for an official clarification if they need one.

>Same is true in may experience in Legal contracts many of which are quite >verbose. For the purposes of ICANNATLARGE.COM I believe that the same >experience will be replicated, and hence any dialects in any of those >languages as well as host of others will be very unlikely save Chinese...

I don't recall anyone suggesting the translation of our bylaws-to-be into all kinds of dialects, and I agree it shouldn't be necessary. I don't speak Chinese myself but I gather the difference between Mandarin and Simplified Chinese is great enough that we might need both.

>> I also hear frequent complaints from fellow-translators that since their consciences don't allow them to turn in inferior translations, they end up putting in many extra unpaid hours to turn "good enough" into "good".
>
>  I suppose your concern here is really a argument of what good is as >compared to good enough...

Exactly. If our organization is to be taken seriously as an international body, we can't get away with being less than professional in presenting ourselves, in any language.

>Although I myself can read a french newspaper without too much trouble, 
>I would in no way call myself bilingual as french as a second language.  >Yet I have noticed on Jefsey's France@large Mailing list, now dormant, 
>that much of the french used there by frenchmen/women was far less that >what you seem to call "Good"...  And that forum was a public forum.  
>Hence I can therefore only reasonable conclude that perhaps you are >overreaching a bit here...

Apples and oranges, Jeff. We were talking about documents that represent an international organization, not quick e-mail messages and forum postings by individuals. 

Most people don't write especially well, even in their own language -- it's a skill you don't necessarily acquire by graduating from secondary school or even a university. Even people who can write well and do so for a living don't take the same care over informal writing as with something destined for print or a Web site. Most of the professional-oriented mailing lists I get consider it poor netiquette to criticize others if the note they dashed off hurriedly contains errors in spelling or grammar. Most of the professional-calibre Web sites I know submit their pages to careful editing but Web forums are places where things are posted "as is" and everyone knows it.

>I am only suggesting that translation software tools are and aid in >reducing the load on this burgeoning organization with proofreading 
>being reduced and turnaround on documents as something that is produces 
>a "Good" translation...  Hence either you misunderstood what I have 
>been saying, or?????

Perhaps. I do know I've been saying that the use of such tools doesn't necessarily reduce the workload, and that careful editing of the results
is likely to be more necessary rather than less so. Whether a translation was produced by a human, a program or a combination of the two, the time to proofread it is roughly the same. Proofreading is the final stage of the process, where the text is checked letter-by-letter and any last-minute formatting errors are caught before the material is published. 

Proofreading is an adjunct to, not a substitute for, editing which is scrutinizing meaning and style as well as form, and which in the case of translation usually involves careful checking of the translated text against the original version.

I am merely recommending that we not be any less careful of our organization's materials in other languages than we would be for our own, or less careful of documents destined for our Web site than for documents destined to be published as an official report.

>> We have amongst us people who speak many different languages at home but write in intelligible English here.
>
>  Hummm?  I thought you stated or have been arguing that JUST "intelligible" is not good???  Or did I misunderstand you on that???

Nope! "intelligible" when writing a second language is more than most people ever achieve, and it took me about twenty years of hard slogging to be able to do it *reliably* in French. 

But "JUST intelligible" isn't good enough for an international association that hopes to be taken seriously, and it's certainly grossly inadequate for am organization which hopes to form a credible international democratic institution with input into policy decisions. 

In fact, it isn't really good enough for a company or individual looking to do business with people in another country, or for a professional communicating with professional colleagues elsewhere. Credibility often depends on presenting oneself well.

>  Yes such minor mistakes in the use of language can cause such problems.  >But those instances are rare...

I only wish they were ... and so do the many people whose marketing campaigns failed.

>  Your contention that "no program extant really understands human >language" is a personal opinion that is not broadly shared and therefore is >of limited value...

That is your opinion. I can only suggest that anyone interested in the issue do a little research on the Web to see what the professional linguists and artificial intelligence people think about the current capacity of computer programs to comprehend human languages. In our time, computers are getting significantly better at complex pattern recognition than they were even ten years ago but they have yet to learn how to understand what is being said in its cultural context, let alone transpose it into the cultural context of another human language.

I'm not saying it will never happen, or that I wish people would stop trying -- just that even the very best efforts in that direction by the people who devote their lives to the subject (which I don't, obviously) have not yet created a single program which genuinely understands the whole of even one language, let alone two in parallel. If you believe otherwise, so be it.

Regards,

Judyth

P.S. especially for Jeff:  Since the members of the Panel all get this list and seem to be reading it faithfully, I agree with Joanna that it's not necessary to make sure they get duplicate postings. Personally, I'd greatly appreciate it if people didn't send me individual copies as well as the list ones -- neither my hard disk nor my access time are unlimited and I read well enough to get the point the first time. Unlike Joanna, I am not disregarding your postings yet but I'd urge you to remember that being rude is not the best way to persuade people you're right about something.

##########################################################
Judyth Mermelstein     "cogito ergo lego ergo cogito..."
Montreal, QC           <espresso@e-scape.net>
##########################################################
"History teaches us that men and nations behave wisely once 
they have exhausted all other alternatives." (Abba Eban)
##########################################################
See the UNESCO OBSERVATORY ON THE INFORMATION SOCIETY!
http://www.unesco.org/webworld/observatory  



---------------------------------------------------------------------
To unsubscribe, e-mail: atlarge-discuss-unsubscribe@lists.fitug.de
For additional commands, e-mail: atlarge-discuss-help@lists.fitug.de