| 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
|
|---|
Lieven Malaise wrote:
paid MT is even useful to uplift translation quality
Interesting. Could you elaborate on how MT can improve translation quality?
| | | | Lieven Malaise
Belgium
Local time: 01:11
Member (2020)
French to Dutch
+ ...
Ice Scream wrote:
Interesting. Could you elaborate on how MT can improve translation quality?
The AutoSuggest function provides you with synonyms and alternative solutions to formulate sentences that you might not have come up with at first because you have developed your own way of formulating things. Seeing many options for every single sentence broadens your perspective and might lead to a better translation.
I have to add that this doesn't count for machine pre-translated CAT integrated texts that are sent to you by one of your clients, which is mostely the case if you are offered an MTPE job (in that case you can't use the Autosuggest function, or at least not in a direct way).
It can also be helfpul for terminology. If you have a very technical term and you can't immediately find a proper translation, one of the MT suggested translations sometimes gives you a hint of where to look.
| | | | | I think you need a reality check | Aug 2, 2022 |
BabelOn-line wrote:
It is about providing what the TAUS calls "good enough" quality. The response I feel I get is a binary choice between noble human translation and horrible MT. Nothing exists in between. You are either or. Black or white.
Which part of the translation effort can we save, then? Getting the punctuation, spelling and syntax right? Running the spellcheck and QA? Checking if MT got the difficult terminology right? Checking if the meaning is the same? Checking that the terminology is consistent?
We still need the cognitive effort to understand the source and the proposed target and compare them. We can't just do half of that.
BabelOn-line wrote:
As an agency, if we don't manage to find a way to process these 3 million words in a way that is acceptable to all parties (i.e. pricing, time spent, rates for translators), it is lost for everyone AFAIK. For us, for our freelancers.
Is this a problem?
I would like a 4-bedroom detached house with a large garden and a swimming pool for the price of a 1-bed terrace in moderate condition. If nobody steps forward to offer a solution, this wonderful order will be lost for everyone.
BabelOn-line wrote:
As you say, we have strong economic headwinds. So it very much looks like it is going to be cutthroat competition for all linguists, no matter how good or experienced they are.
That's what the agencies would have us believe, at least. This isn't quite the reality I see. I happily let others deal with the agencies that want cut-throat competition.
BabelOn-line wrote:
With this in mind, I find the lack of appetite for "giving it a go" – by this I merely mean doing a sanity check, a bit of exploration: give a glance to the MT output, see how it could be handled, if it can be handled at all, see what kind of quality the client may be perfectly happy with – equates to letting these dark times come without a plan B.
This is what some of us have been doing, as you can see if you read our comments. Many of us have reported that no time is saved in many cases. I record my time consumption for all tasks, so I can easily see if MT results in lower time consumption. In most cases, it doesn't. If I spend the same time with and without MT, it would be absurd to offer reduced prices.
BabelOn-line wrote:
I am simply sad i did not find the way to engage correctly with the top of the profession to see if we could rejig our methods a bit.
Please elaborate on how you think we could 'rejig our methods a bit' to save all that time, then.
If you consult a lawyer, accountant, private doctor, architect, notary, etc., do you also expect them to provide semi-automated solutions that can lower your bill significantly? One of the main problems we have in our industry is the flood of semi-professional agencies with a disgraceful obsession with paying as little as possible, not to mention a flood of low-quality providers that are only too happy to work for a pittance, to the effect that our profession must be one of the least respected intellectual professions. I'm not playing that game by joining the race to the bottom, and I don't need to.
We cannot merge our thoughts with a computer, so as long as software cannot deliver a 100% usable translation, you will need experienced translators to go through the machine output.
| | | | BabelOn-line
United Kingdom
Local time: 00:11
English to French
+ ...
TOPIC STARTER | Yes, and I also want a pink unicorn | Aug 2, 2022 |
Thomas T. Frost wrote:
Please elaborate on how you think we could 'rejig our methods a bit' to save all that time, then.
If you consult a lawyer, accountant, private doctor, architect, notary, etc., do you also expect them to provide semi-automated solutions that can lower your bill significantly?
Dear Thomas, do you really believe that notaries don’t use fully automated software to prepare their transactions in 2022? Why do you think most clerical jobs at notaries disappeared in the last 20 years? The very guys who were preparing notarial transactions by hand? Today, the notary asks a series of question straight out of an expert app on his laptop and fills in the answers. This in turn generates the legal draft. They did not lower their rates, but they kept the change and secretarial people lost their jobs in the meantime.
Also, you can't hire an architect, a lawyer or a doctor out of the country. Not so for translators. You, me, everyone is competing with someone somewhere who is hungry for work.
Rejigging methods would start by opening up your mind. I am not even saying it is possible to obtain MTPE of the required quality at an acceptable price, that there is a silver bullet: all I am trying to find is solution to offer this "good enough" level of quality. Maybe it is possible. Maybe it isn't.
You don't want to be part of it and seem perfectly happy with your job? I am honestly glad you are.
You don't even want to hear about the quality of the raw MT output looks like because you know in advance that it will be bad? Fine.
You think that MTPE is of limited use, it does not save time and agencies are not going to insist on linguists using it and that it will never catch up to the point of threatening "by hand" translation? I hope you are right.
I can't help to notice that no one on this thread asked if the aim was to reduce the costs by 10%, 20%, 50% or 80%. What it meant in decrease in the level of quality. What were the expectations/hopes for the increase in output. Or what the subject at hand was.
Rejigging could start here. Before saying “it will never work” (or worse, “this guy is trying to unfairly pressure us on price”), just try to see what is the question at hand.
Your example hinting at the fact I am looking for a penthouse for the price of a one bedroom is striking, but frankly way off target as you clearly have no idea of what I am trying to explore. Dare not say "achieve" here.
The whole point of me starting this post was to explore why there is so much resistance to MTPE – without even looking at the details. Why the "over my dead body" visceral reaction to it.
Answers I received here are often the line of “I am not interested" (a perfectly valid and respectable answer).
Nevertheless, when I start getting hints that I may be dreaming or a bit delusional, I say this: prêt à porter (standard size clothes instead of taylor-made). Personal computers. Music CDs. Fax machines. Airbnb. Amazon. Über. They all happened for better or worse, often for worse. But they happened as the market wanted them. I for myself keep a weary eye on the market.
| | |
|
|
|
Steve Robbie
United Kingdom
Local time: 00:11
Member (2017)
German to English
+ ...
| If I recall correctly... | Aug 3, 2022 |
... you started the thread with the words "I deeply dislike MTPE", which is to say, you pretty much invited the kind of "over my dead body" reactions you got.
As for me, I post-edit quite happily for one of my clients. I wouldn't translate a novel with it, but many of the commercial documents I deal with are quite formulaic, and a half-decent MT definitely gives me a head-start with those. The time saving should not be overestimated, but I'm convinced it does exist.
| | | | Platary (X)
Local time: 01:11
German to French
+ ...
BabelOn-line wrote:
But they happened as the market wanted them.
Is this certain? Is it not the opposite? They have imposed themselves on the market and often with unglamorous methods.
But at the end, that's not the point.
Instead of trying to find explanations for what has none, it would be much simpler to say that post-edition (or any other name) has nothing to do with translation.
It's a different business and those who want to make people believe otherwise (notably certain agencies) are wrong and I doubt their good faith.
| | | | | Please do not deform | Aug 3, 2022 |
BabelOn-line wrote:
Also, you can't hire an architect, a lawyer or a doctor out of the country. Not so for translators. You, me, everyone is competing with someone somewhere who is hungry for work.
A translation into Danish requires a Danish translator regardless of where he or she is physically located. It cannot be provided by a native Indian. I'm not 'hungry for work', but maybe your view of this profession is that we are all desperate for work. Good Danish translators are not selling their work cheap, and I know from clients that they often find it difficult to find good translators. To take another example, I regularly translate UI strings for server software. The agency tells me that the end client in India has assured him that their writers are native-English writers, but they deliver garbled, sometimes incomprehensive texts with awkward and clunky wording full of every mistake in the book. They could not replace a native English translator or writer either.
BabelOn-line wrote:
Rejigging methods would start by opening up your mind. I am not even saying it is possible to obtain MTPE of the required quality at an acceptable price, that there is a silver bullet: all I am trying to find is solution to offer this "good enough" level of quality. Maybe it is possible. Maybe it isn't.
I don't have a closed mind in relation to MT, but practical trials (Amazon and DeepL) have not demonstrated any time saving, except for formulaic legal texts. You don't seem to have any concrete suggestions yourself.
BabelOn-line wrote:
You don't want to be part of it and seem perfectly happy with your job? I am honestly glad you are.
If it doesn't result in reduced hourly payment, I don't have any problem with participating. So if you pay by the hour, as I have already suggested, a suggestion which you ignored, let's get started.
BabelOn-line wrote:
You don't even want to hear about the quality of the raw MT output looks like because you know in advance that it will be bad? Fine.
You ask us for opinions, but you don't seem to read what we are telling you. Many of us have already experimented with MT and have been telling you about our experiences.
BabelOn-line wrote:
You think that MTPE is of limited use, it does not save time and agencies are not going to insist on linguists using it and that it will never catch up to the point of threatening "by hand" translation? I hope you are right.
My practical trials show that it doesn't save time, except in specific cases. This may depend on the type of text and the language combination. It can help with typing, and it can occasionally suggest a useful term. In case of formulaic texts, it can be more helpful.
I don't know if agencies will insist on linguists using it. I don't have a problem with that as long as it doesn't result in reduced hourly payment.
I can see from practical trials that it has a long way to go before it could replace a human translator without the need for human editing, and at its present stage, it is incapable of correcting source errors. Much of the English I get to translate is clearly not written by English natives or native-equivalent writers, and I often have to compensate for that in the translation to avoid producing 'garbage in, garbage out' translations. MT will also not catch logical errors such as when an UI string incorrectly says that the end date must be ahead of the start date. MT is also unlikely to catch phrases that may need to be localised a bit due to national sensitivities. When for example a US brochure suggests that the user could mount his hi-fi equipment on the wall to showcase it to friends, we need to know that such bragging is often frowned upon in Denmark, so it's better to find another reason for the wall mounting. When do you think MT will be able to do all that and other extras that a professional human translator naturally does, and get it right?
BabelOn-line wrote:
Rejigging could start here. Before saying “it will never work” (or worse, “this guy is trying to unfairly pressure us on price”), just try to see what is the question at hand.
I just don't see how you intend it to work, and you don't seem to have any suggestions yourself.
BabelOn-line wrote:
The whole point of me starting this post was to explore why there is so much resistance to MTPE – without even looking at the details. Why the "over my dead body" visceral reaction to it.
I and many others have been looking at the details and have told you about the results. You just choose to ignore what we say. There is no visceral resistance, but MT at its present stage is not the miracle tool some agencies seem to think it is. We are the ones doing the work, after all. We should know.
BabelOn-line wrote:
Nevertheless, when I start getting hints that I may be dreaming or a bit delusional, I say this: prêt à porter (standard size clothes instead of taylor-made). Personal computers. Music CDs. Fax machines. Airbnb. Amazon. Über. They all happened for better or worse, often for worse. But they happened as the market wanted them. I for myself keep a weary eye on the market.
These example all include elements of mass production, and that is something you can't achieve in translation, where every task is unique. As for Uber, as far as I know, the drivers are paid a pittance when you take into account that they have to pay all the vehicle costs.
| | | | Maciek Drobka 
Poland
Local time: 01:11
Member (2006)
English to Polish
+ ...
| I have started accepting MTPE projects | Aug 3, 2022 |
After being a strong oppontent of MTPE for years, I am now willing to accept MTPE projects that offer an attractive combination of [1] fairly easy content (I have found that MTPE is particularly well-suited for UI translations IF the source text is good quality), and [2] a good base rate.
I elaborated on the topic in a recent poll disc... See more After being a strong oppontent of MTPE for years, I am now willing to accept MTPE projects that offer an attractive combination of [1] fairly easy content (I have found that MTPE is particularly well-suited for UI translations IF the source text is good quality), and [2] a good base rate.
I elaborated on the topic in a recent poll discussion: https://www.proz.com/forum/poll_discussion/358333-poll_i_have_started_doing_more_post_editing_of_machine_translation_in_the_past_year_or_two.html#2964683 ▲ Collapse
| | |
|
|
|
| If the client... | Aug 3, 2022 |
will get a good commercial outcome (in the millions) via the MTPE project, £200k is a small investment. Otherwise, post-editing that kind of volume for a small return is just silly.
| | | |
I suppose you can build a workable MT solution for certain types of projects if you have good TMs for such projects and your MT is teachable. Maybe some language pairs have big corpora of actually good translations (for legal texts, e. g.), so people who work in those language pairs can get decent leverage with MT. This is the only explanation I can come up with for some posts on this forum where people claim MT is getting shockingly good (some of those posts being quite aggressive at that). I h... See more I suppose you can build a workable MT solution for certain types of projects if you have good TMs for such projects and your MT is teachable. Maybe some language pairs have big corpora of actually good translations (for legal texts, e. g.), so people who work in those language pairs can get decent leverage with MT. This is the only explanation I can come up with for some posts on this forum where people claim MT is getting shockingly good (some of those posts being quite aggressive at that). I have shared my experience elsewhere, so I won't get into details now. If I were offered a decent rate for doing MTPE, I'd go for it, but there is no way I'll take on an MTPE project at half the translation rate. The 'how' part of my work is none of the customer's business as long as the result is good enough. Oh, and I wonder if MTPE can produce translations in formats supported by CAT so I could open them in mQ or Trados: I did two MTPE projects as part of my full-time work, and having to open two .docx files with the source and target texts was a really sad experience. And the translations were awful, of zero help to me ▲ Collapse
| | | | BabelOn-line
United Kingdom
Local time: 00:11
English to French
+ ...
TOPIC STARTER
Giovanni Guarnieri MITI, MIL wrote:
will get a good commercial outcome (in the millions) via the MTPE project, £200k is a small investment. Otherwise, post-editing that kind of volume for a small return is just silly.
Doubt they would.This is something they would not sell onwards. More something they have to provide. Hence why they want just good enough quality.
| | | | | The only visceral reaction I get is the one to being underpaid! | Aug 3, 2022 |
Personally, this is just a job for me. I've had better paid jobs, more interesting ones, and certainly I've had jobs that were more useful to the humankind, which therefore were a bigger part of my identity. I couldn't care less if I do translation, editing, or MTPE, honestly, as long as the agency/end client pays me well.
What is dirty about MTPE is if someone wants to send me something that requires "retranslation", but they want all the quality of my translation work and they wan... See more Personally, this is just a job for me. I've had better paid jobs, more interesting ones, and certainly I've had jobs that were more useful to the humankind, which therefore were a bigger part of my identity. I couldn't care less if I do translation, editing, or MTPE, honestly, as long as the agency/end client pays me well.
What is dirty about MTPE is if someone wants to send me something that requires "retranslation", but they want all the quality of my translation work and they want to pay me a fraction of the price. I will reject a hefty sum if the hefty sum should be double the amount on principle and just because I can. Nobody gets half off just because they run something through Google Translate or some TM even worse than GT and they now think themselves very clever and want a half-off rate from me.
And of course I will not expect to be paid full rate for a very good TM on a repetitive job with loads of repetitions but you still need to pay even for that. I must read it and check the quality. What is the logic behind paying me more to check a human translation than you pay me to check what a TM, especially trained by someone who isn't me but also even the one trained by me, decided is a repetition? There is no logic to it, it's a scam of a sort, but I am willing to play the game so 25% and even 20% per repetition can and will be part of some huge project that still must pay well enough for me to be part of it.
It isn't the agencies either, it's some end clients. I know for a fact that it wasn't the very good agency I work with, which pays me a very decent rate, that wanted me to give half off on a recent project, it's the end client who had once already sent machine translation to be edited for my human translation editing rate and, when we realised what has happened, the PM was just as upset about it as I was. Then the agency, for reasons known only to them (I suppose trying to do what we all try to do - stay afloat), accepted another job from the same client (in which he basically wanted half off) and I declined it. They haven't stopped sending me anything but the work from that particular client and they were very understanding - as I got the offer late in the day, accepted it without vetting since I trust the agency and they've never tried to pull any stunts on me, then realised what I've accepted and sent a borderline rude email in which I explained what I thought of the end client's practices.
So there's nothing wrong with so many agencies and certainly nothing wrong with those I work with, but some end clients are very much beyond the pale. You want to run things through GT or some TM God knows who made, be my guest. You are still paying me my full rate because it takes me just the same to get that word salad to where it's a translation I can put my name behind. Hey, pay me three quarters of my rate and I will spruce up your word salad to read like a low to OK quality text, but know up front that that's exactly what you are getting and be happy with it.
What is absolutely dirty is to want to send me a GT word salad and pay me half of my usual rate to fix it and expect the same quality you get when you pay the full rate.
I am responsible for the agencies I choose to work with and I assure you I pick very professional people and therefore in my experience it isn't the agencies but some end clients that try and push this scam through and then, once in a blue moon, even a good agency succumbs to the temptation of a nice hefty sum on a high volume job, even though the sum should be double that.
So you see, I do see your point, but please make sure we are talking about the same thing. So many times, and this is one of the permanent flaws of this forum - one of many, people aren't even talking about the same thing and they simply talk past one another. If you are talking about MPTE done properly that's fair enough, but I am very much concerned with MPTE used as "the same quality for a fraction of the price" scam and that is a scam.
What use is a million word job to me if all I am doing is underselling my services for a very long period of time? This might be more tempting for an agency to accept since perhaps the translator's doing most of the work and the agency still gets their half, but there is no upside to this for a translator. If one cannot do well enough not to have to work 24/7 for a pittance, then this person needs to change their profession.
We should also seriously consider that freelance translation might be a part-time job in this day and age. I know this leaves the agencies in a position where a good deal of them will have to start accepting even dodgier MPTE jobs, but this really isn't my problem since I am not going to be the vendor working on them.
And if you ask why the visceral reaction, well, because there isn't enough trust between the translators and the agencies regarding the true nature of any one MPTE job, since perhaps our interests here do not align completely. If I am a part-time translator who refuses to work for less than I deem my absolute minimum descent rate, and you are a full-time agency trying to stay competitive in the market and to do that you need the occasional dodgy MPTE.
[Edited at 2022-08-03 19:27 GMT] ▲ Collapse
| | |
|
|
|
Jo Macdonald
Spain
Local time: 01:11
Italian to English
+ ...
| Are you reading suggestions? | Aug 3, 2022 |
BabelOn-line wrote:
can't help to notice that no one on this thread asked if the aim was to reduce the costs by 10%, 20%, 50% or 80%. What it meant in decrease in the level of quality. What were the expectations/hopes for the increase in output.
I suggested several solutions along these lines using machine translation, how some would work and why others wouldn't, quality expectations and price.
Are you reading what people are suggesting or looking for people willing to do a lot of work for as little as possible?
I'm starting to suspect high quality expectations for very low rates like 2 cents/word, which would explain why you're having trouble finding a solution to your problem. It has nothing to do with MT.
| | | | BabelOn-line
United Kingdom
Local time: 00:11
English to French
+ ...
TOPIC STARTER | I choose to ignore what you say. Or maybe not | Aug 3, 2022 |
Thomas T. Frost wrote:
I and many others have been looking at the details and have told you about the results. You just choose to ignore what we say.
No, you have not been looking into the details. That's precisely my point.
You infer your opinion from previous personal experience, which would be absolutely fine if you had any a notion about the specific job i am discussing here.
Did you ask what this speciific project is about? No.
Are we talking about usual hand-made, gold standards here? No. This is something completely different.
Did you ask if the MT output was good? No (it really isn't bad).
Did you ask if the MT was teachable? No (yes, it is teachable).
Did you ask if have a good TM? No (we do have great TMs).
Did you ask to see a sample of the MT output to get an idea? No.
Let's not get burdened by mere details, shall we not?
Then, let's say the other party is intellectually dishonnest.
Again, you could simply say "not for me", you are perfectly in your right to explain your dislike of MTPE, to tell that in your experience MTPE was awful for you - in some given circumstances, which many other linguists have done. Qualify things. Differentiate.
In a word, you are entitled to your opinions, ... but not to your facts.
Finally, there are a few voices here and there saying they started working with MTPE, they seem to say that there is indeed a learning curve but there are clear benefits.
I think I'll talk to them. May be they will help me clarify my project. If THEY tell me it can't work (after seeing the specifics), I will not ignore what they say.
Kind regards
[Edited at 2022-08-03 14:04 GMT]
[Edited at 2022-08-03 14:05 GMT]
| | | | Tom in London
United Kingdom
Local time: 00:11
Member (2008)
Italian to English
| I know this is niggling, but..... | Aug 3, 2022 |
BabelOn-line wrote:
...... you can't hire an architect .....out of the country. .
Actually, you can.
| | | | | 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 
