Pages in topic:   < [1 2 3 4]
MT, "neural" stuff, and the future
Thread poster: Chase Faucheux

Chase Faucheux  Identity Verified
United States
Local time: 11:43
German to English
TOPIC STARTER
Good points, Kaspars Apr 22

I agree with what you say, but it's also true that in most cases, the quality of a translation is not a life-or-death issue. I still think that reduction in quality due to market forces is not something any of us should be okay with (except for those who happen to benefit from this), and I still think it's important to have translators who are willing to go deep down the rabbit hole to make sure they fully understand the subject matter and that what they're writing in the target language is in f... See more
I agree with what you say, but it's also true that in most cases, the quality of a translation is not a life-or-death issue. I still think that reduction in quality due to market forces is not something any of us should be okay with (except for those who happen to benefit from this), and I still think it's important to have translators who are willing to go deep down the rabbit hole to make sure they fully understand the subject matter and that what they're writing in the target language is in fact how that concept would be expressed in the target language.

Unfortunately, much of the market incentives, especially at the less established level, seem to reward quantity over quality. If it's possible to make more money producing mediocre material that's "good enough", why even try to do better? I don't see MT getting in the way of this trend, but rather exacerbating it.
Collapse


Kaspars Melkis
 

Phil Hand  Identity Verified
China
Local time: 01:43
Chinese to English
Sigh, the job would be so much easier without those pesky clients, eh? Apr 22

Kaspars Melkis wrote:

Overestimation of MT capabilities can be very dangerous because it sends the wrong message which creates false beliefs. The results can be very damaging to everyone involved. We as professionals have a duty to educatate the public about involved risks so that costly mistakes can be avoided.


I do agree with this. A lot of people genuinely believe that Google can do translation now, because the various tech giants keep telling them so. It is a useful corrective to keep reminding the general public, that actually, no, MT won't work for your job. (This requires that we be realistic about what it will work for - gisting of shallow media pieces.)

That said, Dan is right as well. Keep your head down and keep turning in good translations, and even the wayward clients will return to you in the end. It's in the nature of clients that they want to try the shiny new thing, and sometimes you've got to let them try it, then come back to a real translator when they need it done right.


Kaspars Melkis
Maciek Drobka
Michele Fauble
Philip Lees
 

Dan Lucas  Identity Verified
United Kingdom
Local time: 17:43
Member (2014)
Japanese to English
Quality has meaning only in context Apr 23

Chase Faucheux wrote:
"...but if that's the case, then it sure isn't quality."

It isn't? So who draws these arbitrary lines of quality, and where? As freelancers, we decide which clients we want to work with. If we decide that we don't want to work with certain clients, or in certain markets, or for certain rates, that's fine. That's our right. But it is the right of clients to decide how much quality they need, how much time and money they want to spend on translation. Quality is in the eye of the beholder.

"...then those of us who can't tolerate that and actually try to think about how to render it in proper English idiom need to move aside."

You don't need to move aside, but if almost nobody is prepared to buy what you're selling then, realistically, you've already been bypassed. (My view is that most of the time the issue is that freelancers are not shrewd enough or pushy enough to find those who will buy what they are offering.)

Frankly, potential clients don't care about what you or I can or cannot (shouldn't that be "will or will not"?) tolerate. This is a business. If you want only to focus on beautiful language, then concentrate on literary translation and be prepared to tighten your belt a notch or two. The starving artist image has never really appealed to me but I recognise that it's a thing in some circles.

As for myself, sure, I make my translations read as naturally as I can within the constraints imposed by the client, but I have clients who have glossaries, or who prefer certain turns of phrase. I seldom argue with them. If they want stilted English, I'd rather they paid me for it than somebody else. I have bills to pay and mouths to feed, not to mention an upcoming motorbike holiday riding the off-road trails of the Spanish Pyrenees by day and sampling the local tapas by night. All of this costs money.

that makes it that much harder for quality and dedication to stand out. Not impossible, but harder.

Possibly, though I think dedication is a rather nebulous concept for most clients, who tend to be focused on outcomes rather than processes, and I reiterate my point that is is clients that determine quality. In my experience those clients that are prepared to pay for quality can find it pretty quickly. Most clients simply do not want and cannot afford to pay freelancers working at the top 10-20% of the market. But if they do want such translators, and if they perceive you to be competent and reliable, they will come back again and again.

I agree with what Kaspars says about educating clients, but that only goes so far. I think the proportion of clients who just need to have translation "explained" to them is quite small. The real problem is that most clients really do have fixed budgets, and in the short-term it is unlikely that the client can change that, however much we try to educate them. So they get a machine translation, or a translation from somebody who is not a native speaker of the target language, or a translation from somebody who has no understanding of the technical field in question.

We also need to educate translators about what it takes to run a successful business, and about the probability of success in that business, because not everybody will make it. Freelancers need to be realistic about that, and look to themselves rather than blaming CAT, MT, or the shortsightedness of clients.

If you (in the general sense of the word) are not getting solid flows of work, then there is something wrong with the way you are approaching the job. Maybe you're just not in the top echelon of translators, or maybe you have no credible specialisation, or maybe you refuse to use CAT tools, or maybe you're not reliable, or maybe you're a pain to work with.

The projects are out there. What struggling freelancers need to consider is why those jobs are going to other translators, and not to them.

Regards,
Dan


Rachel Waddington
Mr.Q
Kay-Viktor Stegemann
Michele Fauble
Jorge Payan
Gareth Callagy
 

Lincoln Hui  Identity Verified
Hong Kong
Local time: 01:43
Member
Chinese to English
+ ...
A stream of consciousness Apr 23

The segments most affected by MT are the traditional "money-making" ones - legal, medical and technical. These are the fields that truly affect life and death, and the ones that have very few differentiating factors as far as quality is concerned. The rules are already there, and you are either correct or you are not. This is why they are eminently suited to machine translation, and this is where jobs can and will be lost.

Where MT has made zero inroads is when the translation is an
... See more
The segments most affected by MT are the traditional "money-making" ones - legal, medical and technical. These are the fields that truly affect life and death, and the ones that have very few differentiating factors as far as quality is concerned. The rules are already there, and you are either correct or you are not. This is why they are eminently suited to machine translation, and this is where jobs can and will be lost.

Where MT has made zero inroads is when the translation is an integral part of the product being sold, or otherwise delivered. Stuff that you actually want people to read and understand. Which is why the entertainment industry has been all but immune to MT - well, people have tried, and they ended up having to pull their product off the shelves. Bad translations and low rates aren't exactly a new thing, and if anything, the quality of translations in mainstream media has improved massively compared to twenty years ago.

They may not pay terribly high on the surface, but there's enough money in the game that there is a resistance against dropping the rates beyond a certain point. Because much of this stuff is natural language rather than specific terminology, someone who has a strong grasp of language flow can work very quickly, while someone who doesn't will work slowly and make painfully obvious errors. I've had a few of these cases pass through my hands lately.

The notion that translation quality used to be better back in the day is simply wrongthink. What's really happening is that things that previously were not translated are now being translated because reasons, and in some cases the bubbles of false demand are bursting. In other cases there never was a demand to begin with. It's new people deciding they want a translation and balking at the prices they need to pay, not old people who need a translation deciding they don't want to pay anymore. What kinds of websites do you think integrate a Google translate toolbar? In many cases it's academic organizations and other associations - not corporations that have a corporate image to maintain and products to sell. Some businessmen are idiots, but not all of them are, and nobody knowingly plays games with their money. If they haven't commissioned a translation, it's because it doesn't result in net gain for their business, and the ones that commission bad translations are in the same boat; they don't really have a demand for it and are just going through the motions. If there is no real demand driven by legitimate business reasons, it's a bubble, and bubbles burst.

Literally no traditional industry has benefited more from the proliferation of technology, in particular the internet, than translation. It's easier to enter the profession than it ever has been. Whatever threats you think translation is facing from technology, virtually every industry is facing them, and people are in more danger of being competed out of business by new technology in most professions you can think of. Right now, the amount of literature on AI that needs to be translated is outpacing AI's ability to keep up.
Collapse


Chase Faucheux
Kay-Viktor Stegemann
 

Kaspars Melkis  Identity Verified
United Kingdom
Local time: 17:43
English to Latvian
+ ...
MT is not for medicine Apr 23

I don't know about legal and technical, but medicine is definitely not a field where MT has any chance to take over any time soon.

It is exactly because they have to be precise in meaning that no MT system can guarantee.

Also medicine, and particularly pharmacy are industries where process is even more important than final checks. We cannot for practical purposes to check every batch of manufactured medicine for all quality criteria therefore it is extremely important t
... See more
I don't know about legal and technical, but medicine is definitely not a field where MT has any chance to take over any time soon.

It is exactly because they have to be precise in meaning that no MT system can guarantee.

Also medicine, and particularly pharmacy are industries where process is even more important than final checks. We cannot for practical purposes to check every batch of manufactured medicine for all quality criteria therefore it is extremely important to follow every step of manufacturing that are validated and approved by regulators.

Therefore process is important also in translation. One cannot say: I will use MT and edit the output because the result is the only thing that matters. While technically correct, when the result cannot be easily checked, it becomes important that every step is followed precisely. We have long time experience that the current process of translation/editing/proofreading works sufficiently well as long as properly qualified linguists are doing it. We don't know if changing any step, for example, replacing translation with MT, will produce the same results. Medicine is the field where things should not be changed without sufficient evidence.

Another reason is that pharmaceutical texts can be extremely complicated. A simple drug information document will include aspects ranging from therapeutics, pharmaceutical chemistry, statistics, microbiology, as well as require regulatory complicance from at least 4-5 different sources and legal conformance.

Even if the MT output that was reasonably good, checking its conformance with all refrence sources would take as much as time as translating it by hand. In fact, it might be more risky because MT often produces sentences that at the first glance seem fine but only when you analize them deeply, you realize they are misleading and wrong. There is a real risk that revisers will overlook these mistakes that human translators usually do not make.
Collapse


Chase Faucheux
Luca Tutino
 

Chase Faucheux  Identity Verified
United States
Local time: 11:43
German to English
TOPIC STARTER
I guess I get your point Apr 23

But it is the right of clients to decide how much quality they need, how much time and money they want to spend on translation. Quality is in the eye of the beholder.


In some sense, I do see your point. If what you mean is that from a business point of view, "quality" really is about what people are willing to pay for, then yeah, I get it. Perhaps a standard of "English texts should read like they were written by a native English speaker" is too high a standard, and if the client wants less than that, it's only fair to oblige them.

But if what you mean is that quality is some nebulous thing that can't be pinned down, then I disagree. To me, quality is a text that reads like it was written by a native speaker of the target language, in the same register as the original, conveying the same factual and emotional information as the original. If the SL is written in a broken form of the language, filled with misused words and factual errors, then perhaps the translation should read that way to. But if a translator is introducing errors and ambiguities into the text by carrying over SL idioms and making guesses on words with multiple translations (Is German "Land" land, country, countryside, or what?), then that's only "quality" in the sense described above.

Frankly, I think it's absurd that you're assuming I want to focus only on beautiful language. The examples I gave are not bad because they're inelegant -- they're bad because no native English speaker writes or talks that way. Elegance and beauty are beside the point. I've read lots of ugly sentences written by native speakers. I've never read a native speaker who produced nonsense like "With its two settings, this product convinces," obviously a carryover from German syntax and idiomatic use.

Of course if a client provides a glossary, I'm bound to those terms, however bad (glossaries themselves are often prone to the same issues of laziness regarding terminology and usage as anything else), and I do pass up on jobs that don't pay me enough to do the job correctly. I once had to stop accepting jobs from a WWII memorabilia site because while the work was very interesting, the company was unwilling to factor in the amount of research necessary to do the job properly. What's going up on their site now is utter garbage.

And believe me, I know I'm not among the most business savvy folks out there. But I see lots of translators getting big jobs and doing awful work (and now that I think of it, they might well be using MT and post-editing), then I get the privilege of doing the proofread. Because most of the agencies I work for are very fair, they often do agree to pay me more to revise these poor translations after reviewing the changes that need to be made, but it still gets my goat that translators with such low standards get the higher translation rates because they're more focused on quantity than quality. I guess the joke's on me, though.


 

Chase Faucheux  Identity Verified
United States
Local time: 11:43
German to English
TOPIC STARTER
Good points all around Apr 23

Lincoln and Kaspars,

Very good points made. I'd say the fields Lincoln mentioned do face certain pressures because the language tends to be so formulaic. The catch is that the stakes are so much higher in precisely those "formulaic" fields. Most of the work I do is in tourism, marketing, and entertainment -- fields where a need for natural and original language make it harder for Skynet to take over. But none of it is life-or-death, either. So it seems that fields least suitable for
... See more
Lincoln and Kaspars,

Very good points made. I'd say the fields Lincoln mentioned do face certain pressures because the language tends to be so formulaic. The catch is that the stakes are so much higher in precisely those "formulaic" fields. Most of the work I do is in tourism, marketing, and entertainment -- fields where a need for natural and original language make it harder for Skynet to take over. But none of it is life-or-death, either. So it seems that fields least suitable for MT have a lesser potential for disastrous consequences than those in which the language is less "readable", but the content is so serious that the work should not be left to machines.
Collapse


 

Phil Hand  Identity Verified
China
Local time: 01:43
Chinese to English
One area where client (and translator!) education would useful Apr 24

Dan Lucas wrote:

But it is the right of clients to decide how much quality they need, how much time and money they want to spend on translation. Quality is in the eye of the beholder.


Of course this is true... but...

I don't know if this is limited to my pair, but there does seem to be a genuine and widespread misconception among a lot of people - even translators in my pair! - that you can distinguish between "language quality" and "translation quality." Overheard (on an almost daily basis) in the office at my last job: "Oh, for this one the language doesn't have to be great, just get the meaning right and it'll be fine." Probably you know and I know that this is just standard excuse language, made up by poor translators to cover their blushes. But often clients seem to believe it.

A client can certainly decide how much they want to spend, but I think they often don't know what the quality curve looks like in translation. They think that if they spend 70% of the money, they'll get 70% of the quality. In reality, when you spend 70% of the money, you enter a quantum state of uncertainty: if you're lucky, you'll get a decent translation from an underpaid student, but there's a much higher chance that you'll just get dreck.

There are two things that I'd like clients to know:
(1) If a translator (machine or human) is making "language mistakes," then they are *definitely* also making meaning mistakes.
(2) Translation quality is surprisingly all-or-nothing. If a document is only 50% readable, no-one will read it, and so you lose 100% of the value.

While the customer has a perfect right to set their own budget, I think they also have a right to know what they're getting for their money, and understanding these points would help customers make good decisions.


Chase Faucheux
Kaspars Melkis
Rachel Waddington
Philip Lees
Luca Tutino
 

Dan Lucas  Identity Verified
United Kingdom
Local time: 17:43
Member (2014)
Japanese to English
Well-intentioned, but will it have any effect? Apr 24

Phil Hand wrote:
While the customer has a perfect right to set their own budget, I think they also have a right to know what they're getting for their money, and understanding these points would help customers make good decisions.

I see your point, but are clients who take such a casual approach to translation ever likely to be the kind of client who pays the rates necessary for a good translator?

Serious people take every part of the product seriously, and consider the trade-offs carefully. They don't say "Aah, you know what? We'll get Xiaxue* in accounting to do a quick translation. She did a 3-month homestay in Montana - it'll be fine. Let's just whack it in there and ship." By definition, that's not a good client. Things might change, but how long can you wait? Why not just look for the better clients?

(I suspect factors specific to the culture of the source language are also relevant here. Good Japanese clients tend to be very careful, but are often extremely rigid. Other cultures may have a more... uninhibited attitude to translation.)

To put it more succinctly, I am not sure that you can turn a pig's ear into a silk purse, however much education you bring to bear. On the other hand, I can't see how it does any harm, other than maybe wasting some time.

In a way, I'm as guilty of attempts at education as anybody, as in my translations I leave a steady trickle of comments on usage, terminology etc. I'm not sure it changes the minds of my clients, but it pre-empts the "Did he just overlook this, or was it a deliberate omission?" on the part of the proofreader. At least they can see the thought process.

The point, however, is that these are clients I have already won, and who therefore are prepared to pay for what they perceive to be quality. I'm happy to do this for the birds in the hand. Birds in the bush are a different thing again...

Regards,
Dan

*Not her real name

[Edited at 2019-04-24 07:59 GMT]


 

Philip Lees  Identity Verified
Greece
Local time: 19:43
Member (2008)
Greek to English
Level of information Apr 24

Kaspars Melkis wrote:

I don't know about legal and technical, but medicine is definitely not a field where MT has any chance to take over any time soon.


I agree with this, and not just because most of my translation and editing work is in the medical field.

I think it's possible to make a distinction here. If we look at the areas of creative thought in which AI has outperformed human beings, notably chess and Go, both are "full information" games, by which I mean that there is no information about the state of the game that is not available to the player: all is visible on the board.

This is not true for all games. Bridge, for example is not a "full information" game because at the start of each deal each player sees only 13 of the 52 cards and has no information about how the rest are distributed among the other three players.

Perhaps we could apply this distinction to translation. In some cases, all the necessary information may be included in the source text, plus any accompanying instructions, glossaries, etc. This category could include straightforward lists of components, or simple instructions that follow a standard format. These tasks could be handled adequately by AI at its current level, or in the near future.

The other category includes source texts that do not contain all the necessary information. In these cases, the translator's knowledge and experience, extending way beyond the task at hand, need to be brought to bear in order to achieve a good translation.

Some medical translations belong in the former category above, but most of the ones I work on do not. I have seen how MT deals with these texts and the result generally consists of a reasonable approximation for a while, followed by a horrendous howler. Then there are other more subtle errors that would only be apparent to someone familiar with the specialty: i.e. the text is not technically incorrect, but fails to follow specialist usage.

Beyond that, medical writing quite often assumes a level of knowledge on the part of the reader, knowledge that is necessary to resolve ambiguities in the text.

All the above considerations apply equally to other fields, of course.

I have no doubt that these problems will eventually be overcome and that MT will be able to produce competent translations of these "non-full information" texts.

But I don't expect to see it in my lifetime.


Kaspars Melkis
Gareth Callagy
 
Pages in topic:   < [1 2 3 4]


There is no moderator assigned specifically to this forum.
To report site rules violations or get help, please contact site staff »


MT, "neural" stuff, and the future

Advanced search







Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

More info »
Anycount & Translation Office 3000
Translation Office 3000

Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search