| Pages in topic: [1 2 3 4] > | MTPE is a dirty word... but client will not pay the rates for full human translation. Thread poster: BabelOn-line
|
|---|
BabelOn-line
United Kingdom
Local time: 15:15
English to French
+ ...
I am a translator. And I run a small agency.
This topic has no other purpose than telling who will listen about the impossible position I am in when it comes to MTPE.
As a translator, I deeply dislike MTPE. You Prozians can probably guess why. The syntax of a sentence one always rejig must - says the Yoda inside the MT.
Even when this is not the case, you have to check every sentence line by line to catch mistakes. It is exhausting. Depending on the subje... See more I am a translator. And I run a small agency.
This topic has no other purpose than telling who will listen about the impossible position I am in when it comes to MTPE.
As a translator, I deeply dislike MTPE. You Prozians can probably guess why. The syntax of a sentence one always rejig must - says the Yoda inside the MT.
Even when this is not the case, you have to check every sentence line by line to catch mistakes. It is exhausting. Depending on the subject (and language combination), MTPE can still save you some time, but it is hard to tell what kind of boost you will actually get before you start the project. So this makes it hard to quote.
As an agency, I would not say I like MTPE, but it is increasingly clear that clients are becoming aware that "MTPE is a thing" and they want their share of the savings. This is compounded by the fact most clients have no yardstick to assess quality, be it MTPE or handmade. Clients also know about the ability of TM to leverage repeats and fuzzies. There is little place to hide.
One of my clients has 2 to 3 million words in store they'd need translated. Even with a super-thin agency margin, a proper handmade specialist translation would cost well over 200k€.
So they don't press the button.
The general view from my (professional, qualified) linguist colleagues is very much along the lines of "tough luck, if they don't have the budget, look for clients who do, we can't translate on the cheap".
As a translator, I respect this standpoint.
No serious linguist seems to be willing to explore the client's quality expectations. It is hand-stitched, top-of-the-line translations or nothing. It is a matter of professional pride. They put their hallmark on their work. I can't blame them, I have always done the same.
The alternative – trying your luck with unqualified or less qualified linguists – is, of course, not worth doing as you hardly add up much value to MTPE.
As the guy who needs to get business in (and incidentally feed this work to our freelance linguists), I am despairing. There is potentially a thousand days worth of work waiting. There is not even a possibility to see how we could tackle this job differently. Maybe see what the MT looks like and assess? Maybe see how we could preserve the translators' hourly earnings while completing jobs quicker?
It very much looks like they are two irreconcilable trends. Pro linguists who don't want to dumb down their job and quite rightly want to maintain their income. Clients with increasing volumes to translate but not willing to pay the "hand-stitched" price.
There is now a concentration trend amongst agencies and I am worried that the profession will be more and more ruled by large operators who (usually) don't give much consideration to quality because it is all price-driven, shareholders want their pound of flesh. This cannot be a good outcome for the real, quality-loving pros.
I'd very much like to have your opinion as I'd like to go forward in the fairest possible way, both with my linguists and my clients. ▲ Collapse
| | | | Joakim Braun 
Sweden
Local time: 16:15
German to Swedish
+ ...
BabelOn-line wrote:
No serious linguist seems to be willing to explore the client's quality expectations. It is hand-stitched, top-of-the-line translations or nothing. It is a matter of professional pride. They put their hallmark on their work. I can't blame them, I have always done the same.
I can't quite believe that it's impossible to recruit decent translators for MTPE. There are plenty of people here on Proz who are not MTPE-averse and who are skilled pros. (Personally I wouldn't touch it, finding it boring and irritating, much more so than editing a bad human translation.)
BabelOn-line wrote:
It very much looks like they are two irreconcilable trends. Pro linguists who don't want to dumb down their job and quite rightly want to maintain their income. Clients with increasing volumes to translate but not willing to pay the "hand-stitched" price.
That sounds right. I believe AI will soon (2-3 years) get good enough to allow mediocre translators to do pretty sharp MTPE. The rest of us will need to improve quality to capture the top market, go into some other business or just grind it out.
BabelOn-line wrote:
There is now a concentration trend amongst agencies and I am worried that the profession will be more and more ruled by large operators who (usually) don't give much consideration to quality because it is all price-driven, shareholders want their pound of flesh. This cannot be a good outcome for the real, quality-loving pros.
That trend has been noticeable for a decade. But the boutique agencies who deliver top quality are still there and for some niches I don't think they need to worry - the market isn't going away soon (regulatory, financials, sales material, medical...).
If I found myself dealing with a lot of MTPE as a translator or agency owner, I'd do my best to quit this business. Not necessarily for love of hand-stitching (though there's some of that!), but because it wouldn't be any fun.
[Bearbeitet am 2022-08-02 11:23 GMT]
| | | |
Sometimes it is easier to translate from nothing since MTPE can mislead everything
| | | | | This world is all about benefits | Aug 1, 2022 |
BabelOn-line wrote:
There is potentially a thousand days worth of work waiting. There is not even a possibility to see how we could tackle this job differently. Maybe see what the MT looks like and assess? Maybe see how we could preserve the translators' hourly earnings while completing jobs quicker?
1000 days of reviewing MT sounds like a nightmare. Most translators want to translate, not spend over two years editing.
However there are MTPE specialists out there, so instead of trying to sell your current translators on MTPE, why not recruit some specialists? Or would-be specialists looking to get some experience, or talented newbies? It's more of a gamble compared to using your seasoned translators, but fortune favours the bold and all that.
Alternatively, show your translators the benefits of taking on this project. Maybe you should take a sample, run it through the MT and edit it to the client's satisfaction so you can check the time saved (if any) and the profit made. That way you'll be able to make a more convincing case, at least in your language pair. "Make $200 an hour!" sounds a lot better than "Spend 1000 days reading machine translations!" right?
[Edited at 2022-08-01 23:40 GMT]
| | |
|
|
|
BabelOn-line wrote:
As an agency, I would not say I like MTPE, but it is increasingly clear that clients are becoming aware that "MTPE is a thing" and they want their share of the savings.
When automobile manufacturers decided to read and optimize the electronic controls of their engines digitally, garage owners had to buy expensive equipment to do something they did by hand beforehand.
Do you think garage owners started charging their customers less after their investment because they now have support from a computer system?
As professionals, they still remain responsible for interpreting and applying the suggestions made by the computer system. That is their real added value.
Isn't our situation as professional translators pretty much the same?
| | | | Tom in London
United Kingdom
Local time: 15:15
Member (2008)
Italian to English
BabelOn-line wrote:
......There is potentially a thousand days worth of work waiting...
That's 1000 days of soul-destroying work, correcting MT, that nobody wants to do.
So you could say that MTPE is killing its own market.
| | | | | Split it up, Scotty | Aug 2, 2022 |
Tom in London wrote:
That's 1000 days of soul-destroying work, correcting MT, that nobody wants to do.
So you could say that MTPE is killing its own market.
They will most likely distribute it among many freelance translators. This is also very good for consistency. (NOT!)
| | | | Jo Macdonald
Spain
Local time: 16:15
Italian to English
+ ...
| Does the client want free, understandable, or quality? | Aug 2, 2022 |
You can get all of the above from MT.
If your client only has a very small (or no) budget to translate millions of words, MT is free, end of story. You get a fast, free, laughable translation.
If they want the translation to be understandable, they need a budget for post editing.
How much that costs depends on:
1. the quality they need
2. how easy to understand the source is
3. how good the MT engine is
4. how good the linguist is at... See more You can get all of the above from MT.
If your client only has a very small (or no) budget to translate millions of words, MT is free, end of story. You get a fast, free, laughable translation.
If they want the translation to be understandable, they need a budget for post editing.
How much that costs depends on:
1. the quality they need
2. how easy to understand the source is
3. how good the MT engine is
4. how good the linguist is at post editing
5. the linguist’s rate
If the source text is easy to understand, a good MT engine like DeepL can save a lot of time and if your client only wants to understand what the source says (light-PEMT, the translation just has to be understandable) a reasonable rate for this would be about half my rate for a quality translation, so about €0.05/word.
If your client wants a publishable translation ready to be presented to the public as an example of the quality they provide it's going to cost about twice that because that's how long it takes. If they want to use MT because they think it costs less, even a good MT engine like DeepL won't save that much time because the actually typing is just one part of the job and MT adds errors like terminology inconsistency, out-of-context translations, etc. The term checking and QA parts of the job take longer if a machine is doing the typing.
If the source is easy to understand, the MT output probably won't be that bad and will save some time, but instead of 50% less time for Light-PEMT it will save about 10% for Full-PEMT, so my rate for that would be about €0.09/word.
If the source isn’t very easy to understand or the MT engine isn't much good, using MT to try and do a good job probably won't save any time at all, and it'll just make the job more problematic and prone to error.
Full-PEMT will not be as good as a translation a pro can produce without being obliged to use MT, but if the client's quality requirements are lower you can save time and money using a good MT engine.
The above all depend on giving the person doing the work a reasonable deadline and enough time to do the job.
One way some agencies cut translation rates using PEMT is by calculating a huge amount of edited words per hour paid at your hourly rate; you probably can't even read that many words all day let alone correct them. This works out at a rate of about €0.02/word.
Imo in this case the client just wants the agency, who just wants the translator, to sign off for the bad job they've set up and take the responsibility no one wants to take for potentially putting someone's life at risk. Personally I want nothing to do with this.
So
- If the client has no budget, just tell them to Google it and use the MT output in an amusing context that's obviously not important in terms of quality.
- If they have half the budget for a pro translation and the source is easy to understand, Light-PEMT can save time and money if all they want is to be able to understand what the source text says in another language.
- If they want a quality translation, they need the budget for a pro linguist and the time required.
Most people understand MT is not great quality, so when quoting a PEMT job if the client wants great quality it’s our job to explain why using MT will not really save time or money.
They might try to get a cheaper deal by repeating the words “Machine Translation” “Machine Translation” but deep down they know a good job is going to cost more.
[Edited at 2022-08-02 09:07 GMT] ▲ Collapse
| | |
|
|
|
Joakim Braun
Sweden
Local time: 16:15
German to Swedish
+ ...
Jo Macdonald wrote:
If the source text is easy to understand, a good MT engine like DeepL can save a lot of time (...)
It's not necessarily about "easy to understand". It's about:
a) The corpus of bilingual text, relevant to the context, available to the MT engine.
b) The correlation of the terms with each other and with the context. Specialized language (in financial reports, for instance) has a lot pf phrases that tend to occur together, helping MT to identify context vocabulary and choose the right term among a multitude of possible translations and (apparent) synonyms. MT is better at narrow-context translation and worse at general, "open" translation (always provided there's a good corpus).
The last point is very obvious from Amazon country pages run through MT - the corpus is way too small for the thousands of specific products and their descriptions, so MT has to pick some random term translation. Then these "translations" show up as references here on Kudoz...
[Bearbeitet am 2022-08-02 09:43 GMT]
| | | | Lieven Malaise
Belgium
Local time: 16:15
Member (2020)
French to Dutch
+ ...
Joakim Braun wrote:
The last point is very obvious from Amazon country pages run through MT - the corpus is way too small for the thousands of specific products and their descriptions, so MT has to pick some random term translation.
Remarkable, because these kinds of translations are offered massively (hundreds of thousands of words on a very regular basis) as MTPE at ridiculously cheap rates. I used to work on some of them before MTPE took these things over. I wonder who takes these jobs, because they are rather difficult (because of the constantly changing terminology and a lot of short texts) and there is absolutely no way that you can earn a decent living by doing them.
| | | |
One-time strategic spending of 200k€ to enter a market potentially worth millions/billions of euros yearly is nothing - as long as it's not me spending them.
The client knows machines work with switches and don't care about volumes, so they see little point in using their internal resources to classify content according to a set of criteria in-house/external, user/sales material, critical/social media, etc.
But the client should also know (or be told) that human translation... See more One-time strategic spending of 200k€ to enter a market potentially worth millions/billions of euros yearly is nothing - as long as it's not me spending them.
The client knows machines work with switches and don't care about volumes, so they see little point in using their internal resources to classify content according to a set of criteria in-house/external, user/sales material, critical/social media, etc.
But the client should also know (or be told) that human translation and PEMT don't look alike. Like incompetent human translation and serious human translation. Like using custom MT engine or public DeepL/GT. Customer-facing people have a duty to explain to them that you don't make a silk purse out of a sow's ear. And custom MT is far from free.
For their own sake to save money, I'd request them to analyse the content : does everything really needs to be translated? Who's going to read 10kpages (or a 1-m-high pile of standard single-sided pages)? Categorize those 2-3Mwords per quality level required. The client would likely be prepared to pay more for their external e-learning material or sales brochures than for their internal memos or e-mails.
Then apply Jo's segmentation (although I don't believe in "full MTPE", which is asking for trouble).
"Resource"-wise, I'm sure many young people in the profession are eager to become MTPE specialists, because the literature (sponsored by large operators) tells them everywhere that there will be no other way to survive as a translator in the short term.
Have them believe it's the opportunity of a lifetime. If they're young enough, a year or two of MTPE are unlikely to irreversibly affect their brains.
Philippe ▲ Collapse
| | | | BabelOn-line
United Kingdom
Local time: 15:15
English to French
+ ...
TOPIC STARTER | So, what's the answer? | Aug 2, 2022 |
For the record, 2 to 3 million words is the total available wordcount for over 30 projects, all requiring different specialist knowledge or a sound understand of the topic at hand – and probably 30 different linguists.
There are indeed MTPE specialists around (and I personally see MTPE as a valuable skillset), but the jobs in question require a sound knowledge of each topics - not necessarily at PhD level, but still. This is why we can't merely rely on "generic MTPE linguists". ... See more For the record, 2 to 3 million words is the total available wordcount for over 30 projects, all requiring different specialist knowledge or a sound understand of the topic at hand – and probably 30 different linguists.
There are indeed MTPE specialists around (and I personally see MTPE as a valuable skillset), but the jobs in question require a sound knowledge of each topics - not necessarily at PhD level, but still. This is why we can't merely rely on "generic MTPE linguists".
For the first two jobs we processed, our three specialist linguists created translations that were in my view far superior to the source (ambiguities were cleared, many mistakes corrected, style was more fluid, etc).
Sterling job. Client was bowled over. Big thumb up. They loved it.
They did not place any further order.
We made a dry run and contacted "non-linguist specialists of the topic", who would act like proofreaders in the target language. Maybe with some luck, they would catch the MT inconsistencies and ensure the terminology was correct. A sort of sanity check. The tryout we did with a PhD student in Economics from a good UK university proved unsuccessful.You need the linguist instinct to detect when MT has gone off the rails.
And this is the rub:
1/ unless one can think of a better way to do this, we need to work confidently with seasoned pros linguists who know the topic at hand - or at least "know what they don't know" and research terminology
2/ the client expectation in terms of quality are contradictory: given the massive volume at hand, they will not order unless we can drastically reduce the rate per word. They do not care if the syntax of the translation is clunky. However, the expectation is the final output should be “factually correct” and “good enough” - as defined by TAUS.
3/ seasoned pro linguists are usually not interested in exploring ways of reducing the costs. They usually dislike MTPE (and again, I fully understand why). They are specialists: why would they shoot themselves in the foot by getting a smaller portion of the pie - even if we maintain their hourly rate?
So may be this is insoluble. Let's simply forget about millions of words.
Someone I know, who worked for one of the very largest multinational agency (as an IT guy), was telling me how English into French translations were carried out initially in Puducherry, India, on the strength this is a former French trading post and French is still spoken there (likely as a third of fourth language after Tamil, Teluglu, Hindi, Malayalam, English, ...). Files were then sent for a quick check in Morocco. Final check were carried out in France by unpaid interns. All this in strict accordance of ISO req's, of course.
This is anecdotal. However, I do not believe such agencies give a damn about our professional pride and taste for quality. We have all seen it: they will promise top quality to the client and pick the lowest bidder. with the concentration, this is where the bulk of the market goes.
They may not get the fabled 3 million words job either – as they price high.
Joakim Braun wrote "That sounds right. I believe AI will soon (2-3 years) get good enough to allow mediocre translators to do pretty sharp MTPE. The rest of us will need to improve quality to capture the top market, go into some other business or just grind it out."
Problem is, top market is not expansible. It is indeed narrowing to a few specialities like legal/pharma/tech. AI is indeed progressing. Does not look too good, if you ask me.
I can't tell what the solution is, but there could be a way. Good linguists could become project leaders. They would work fewer hours on MTPE while maintaining their hourly rates? Maybe they could supervise teams of proofreaders. Or other MTPE linguists. We have not cracked it yet. we would love to engage dialogue.
As an agency, we fought for quality – and the utmost respect for our linguists, in terms of consideration, deadlines and pay – for the last 20 years and we may well be flogging a dead horse. I refuse to quit. ▲ Collapse
| | |
|
|
|
Tom in London
United Kingdom
Local time: 15:15
Member (2008)
Italian to English
| What literature is this? | Aug 2, 2022 |
Philippe Etienne wrote:
..... I'm sure many young people in the profession are eager to become MTPE specialists, because the literature (sponsored by large operators) tells them everywhere that there will be no other way to survive as a translator in the short term
What literature? I've never seen anything that says what you report.
| | | | Samuel Murray
Netherlands
Local time: 16:15
Member (2006)
English to Afrikaans
+ ...
| Not all translators | Aug 2, 2022 |
Tom in London wrote:
That's 1000 days of soul-destroying work, correcting MT, that nobody wants to do.
Not all translators enjoy a challenge. Some of us are perfectly happy doing things like proofreading phone books. Most of what I do might seem terribly boring to some of my colleagues, but I don't mind it. 1000 days of MT checking would be all right for me, if the content is interesting and if the money is right.
| | | |
I have a solution: don't translate those millions of words. Simple. End of story. There is a cost to everything, and if they wanted to build a new office building for 20,000 employees, they wouldn't get it for the price of a 10,000-employee building, or it would be seriously shoddy quality that would probably never pass any inspection.
If there isn't a solid business case for translating those millions of words, then it would make little sense to do it. It's easy to make 'if only re... See more I have a solution: don't translate those millions of words. Simple. End of story. There is a cost to everything, and if they wanted to build a new office building for 20,000 employees, they wouldn't get it for the price of a 10,000-employee building, or it would be seriously shoddy quality that would probably never pass any inspection.
If there isn't a solid business case for translating those millions of words, then it would make little sense to do it. It's easy to make 'if only reality wasn't so-and-so' business cases, but they are not business cases but merely fantasies having nothing to do with sound business principles. The rest of us have to adapt our spending to our budgets. Why do end clients like that think they can get things for half the price?
I don't dislike technology that can make me more productive or increase the quality or both, and MT can sometimes assist, particularly with formulaic texts such as T&Cs, but in most cases, I don't see any time gain from using MT, so there is no saving to pass on to a client. I can perhaps save some typing when the MT result is exactly or nearly what I would have written anyway.
Let them sort out their millions of words themselves. They are mirages. We can't base any serious business on that. ▲ Collapse
| | | | | Pages in topic: [1 2 3 4] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » MTPE is a dirty word... but client will not pay the rates for full human translation. | Trados Business Manager Lite | Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
| | Protemos translation business management system | Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!
The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
