| Pages in topic: < [1 2 3] > | Response to Post-Machine Translation Editing Request at 50% Discount of Standard Rates Thread poster: Dylan J Hartmann
|
|---|
| There is no such thing as PEMT OR LPEMT | Oct 1, 2014 |
Giovanni Guarnieri MITI, MIL wrote:
on the quality of the PEMT (and I have't seen any good quality PEMT until now)... but even so, any productivity gain would be compromised by working in a new tool... at 50% of your rate...
[Edited at 2014-10-01 12:32 GMT]
We talked about this in a few other posts, but I will say here that the concept of "post-editing machine 'translations'" is something I find completely wrong. These are not translations done by a language expert, this is simply word- or phrase matching based on algorithms. The result is horrendous, and I challenge anyone to show me an acceptable text in which grammar, syntax and correct terminology play a significant role - which is what matters in "human" translations to meet the objectives of professionally thinking clients - that would be acceptable for "post-editing."
Just looking at "MT-texts" and accepting that these already are translations is just a joke.
The way these words are presented is never without mistake (or, if you want, put a percentage on it but I don't see it approach anything that could be acceptable with regard to the translation of a coherent original text) and can never be simply accepted as "accurate" without checking the original text. And style is a whole other story. Do you see where this is going? On top of that, it messes with your head. It's like being bombarded with wrong syntax, grammar and words on a continuous basis which might either confuse your own way of thinking and of constructing sentences or even impair it, especially if you keep doing this sort of thing.
MT can never be the starting point in order to arrive at a good translation, and therefore the term post-editing is simply wrong. You need to start with the original text. It's about what YOU do, not what a MACHINE did before you get to it. What the machine did before you get to "continue" is in no way acceptable as a basis and trying to drive the price down for the work that is involved if you were to accept such premise - that is accept MT as translation - and to accept such scheme - is simply wrong, to put it mildly. IMO. And don't even say LPEMT.
If the text was "pre-translated" by a CAT tool on the basis of a TM, the term MT is also false. I would only accept pre-translated texts as "translation projects (not as post-translation editing) if they were based on TMs that were "created" by humans during previous human translations. But in this case, you also need to look again at the original text to assess how good or bad the "pre-translated" text is. I can't just accept what the CAT tool with its TM automatically produced. But this is not what is meant by MT. MT stands for machine translation, and I reject that concept in its entirety.
HTH
[Edited at 2014-10-01 16:55 GMT]
| | | | Chien Nguyen 
Vietnam
Local time: 23:15
English to Vietnamese
+ ...
| I will never take it | Oct 1, 2014 |
I used to edit for very poor quality translations produce by human being, after the first page I decided I'd better to translate from the source text rather than editing that poorly translated text. It save my time and release me from ethical headache.
I am working in the pair of English - Vietnamese.
| | | | Jacques DP
Switzerland
Local time: 18:15
English to French
| This is ridiculous because... | Oct 1, 2014 |
...machine translation is freely accessible. Anyone can freely and instantly run the translation through Google Translate before starting work if desired. (Of course, it is useless.)
Therefore their offer is a complicated way to say they will pay a very low rate for the job. My guess it that the offer selects unsuccessful translators looking for a reason to accept a very low rate.
| | | | Tom in London
United Kingdom
Local time: 17:15
Member (2008)
Italian to English
| Yes, but who on earth wants a bad translation? | Oct 1, 2014 |
Jacques DP wrote:
...machine translation is freely accessible. Anyone can freely and instantly run the translation through Google Translate before starting work if desired. (Of course, it is useless.)
Therefore their offer is a complicated way to say they will pay a very low rate for the job. My guess it that the offer selects unsuccessful translators looking for a reason to accept a very low rate.
Yes, but what kind of person is it who is trying to save pennies on getting cheap translations? Surely the documentation they require can't be in connection with anything serious?
For example, I can't imagine that a manufacturing firm wishing to expand its business by exporting to new countries would be trying to cut corners by going for cheap, badly done translations.
[Edited at 2014-10-01 16:26 GMT]
| | |
|
|
|
Patrick Porter
United States
Local time: 12:15
Spanish to English
+ ...
| The term "post-editing" is part of the problem | Oct 1, 2014 |
Bernhard makes some good points as to why there is probably no such thing as "post-editing". It's still translation no matter how you look at it. I'm a little skeptical about the notion that there is going to be a market for final translations where quality is so inconsequential that it will only take light "editing" of MT output. Perhaps that will be the case for texts that are less important or that have a very short shelf life. In any event, what seems to be happening right now is that so... See more Bernhard makes some good points as to why there is probably no such thing as "post-editing". It's still translation no matter how you look at it. I'm a little skeptical about the notion that there is going to be a market for final translations where quality is so inconsequential that it will only take light "editing" of MT output. Perhaps that will be the case for texts that are less important or that have a very short shelf life. In any event, what seems to be happening right now is that some LSPs are trying to implement MT in their workflow to save costs, but asking for an unrealistic savings by calling it "post-editing" instead of what it really is: plain old computer-aided translation. I've received a few offers like this myself recently and turned them down because the potential productivity gains did not seem to correspond to the discount being demanded. If a client wants to pay 50% of what I normally charge, then the use of their MT (possibly combined with lower expectations about the end quality) would have to at least double my productivity.
For my part, I'm not as anti-MT and in fact use it in my own work. It can increase my productivity and in some cases allows me to accept a slightly lower rate (never as low as 50%, though), staying profitable while being slightly more competitive. I think it's because I only stick to using MT models trained with my own TMs or those provided from my clients. I never use Google Translate or Microsoft Translator as a basis to auto-populate the target text, because they are too general-purpose and it ends up being better to just translate from scratch. That's why I'm open-minded about the idea of client-specific MT. If they didn't just run it through Google Translate and instead have their own domain-specific MT trained with high-quality, on-point translations, then it might be worth considering. And I disagree that it's impossible to end up with a high-quality end translation if you start with MT. It probably is impossible if you only want to lightly "post-edit" it. But with the proper approach, I can and do continue to provide very high-quality end translations whether I use MT or not, just I've done for the past 15 years as a translator. ▲ Collapse
| | | | | It's worth noting | Oct 1, 2014 |
that this text is for a skincare company. All the texts I've ever done for this industry are marketing copy, for which MT is wholly unsuitable.
Machine translation is for getting the gist of a document. Human translation is for finely crafted prose that persuades people to buy things.
| | | | Jacques DP
Switzerland
Local time: 18:15
English to French
| Mental pollution | Oct 1, 2014 |
Tom: No idea.
Another thing to mention is the mental pollution that one must endure in such jobs (assuming one really starts from the MT output). It's a Ted Kaczynski's (Unabomber) nightmare, the human being reduced to a computer aid. Computer's vomit in your eyes and mind all day long, that must hurt.
| | | | | Don't feel kissed by the gist | Oct 2, 2014 |
philgoddard wrote:
Machine translation is for getting the gist of a document. Human translation is for finely crafted prose that persuades people to buy things.
We shouldn't and can't expect MT to get the gist of things. That's my opinion. If anything, we can expect it to confuse things and use wrong terminology, bad syntax and bad grammar. MT might relay intended meaning with regard to some simple sentences or some standard phrases, but then it'll veer off again completely, no matter if there are simple or complicated sentences in the OT. The problem is that there is no "thinking" of humans involved, a machine can't "get" it. Only humans do.
I know that there's plenty of websites and some search engines that "offer" versions in other languages, "arrived at" exclusively by MT. I don't condone this practice. I was never impressed by any business that uses it or the MT text per se.
In order to get a professional translation and to help fulfill my clients' professional objectives, I don't rely on MT for the "correct" gist of a text in another language for which I then would do "some" post-editing. Just to make that clear.
[Edited at 2014-10-02 00:07 GMT]
| | |
|
|
|
jyuan_us
United States
Local time: 12:15
Member (2005)
English to Chinese
+ ...
| I think it is not the issue of whether MT is good or bad | Oct 2, 2014 |
philgoddard wrote:
that this text is for a skincare company. All the texts I've ever done for this industry are marketing copy, for which MT is wholly unsuitable.
Machine translation is for getting the gist of a document. Human translation is for finely crafted prose that persuades people to buy things.
The answer to that question is apparent. The key issue here is that the company want a translation of the quality of a human translated piece for about half of the rate regularly paid to a translator. There is a word called "exploitation".
| | | | Phil Hand
China
Local time: 00:15
Chinese to English
| There are more things under the sun... | Oct 2, 2014 |
Tom in London wrote:
For example, I can't imagine that a manufacturing firm wishing to expand its business by exporting to new countries would be trying to cut corners by going for cheap, badly done translations.
And yet I live in a country that makes a fair part of its GDP from manufacturers who try to cut corners going for cheap, badly done manufacturing...
jyuan wrote:
...The key issue here is that the company want a translation of the quality of a human translated piece for about half of the rate regularly paid to a translator. There is a word called "exploitation".
Yep. If they're a good client, they'll try it once, find that it doesn't work, then go back to buying translation properly. If they're a bad client, they won't learn so quickly.
| | | | Tom in London
United Kingdom
Local time: 17:15
Member (2008)
Italian to English
Phil Hand wrote:
And yet I live in a country that makes a fair part of its GDP from manufacturers who try to cut corners going for cheap, badly done manufacturing...
Ah yes - the iPhone. I was forgetting. Do you find that Chinese manufacturers rely a lot on MT?
[Edited at 2014-10-02 06:21 GMT]
| | | | Samuel Murray
Netherlands
Local time: 18:15
Member (2006)
English to Afrikaans
+ ...
| Well, they do offer a test | Oct 2, 2014 |
DJHartmann wrote:
Translator must be willing to work with our internal CAT tool - XTM or COACH (training will be provided); must be willing to complete a test (free-of-charge) on a machine-translated (standard completion time is an hour).
So, it's quite simple actually: just do their test and see if their MT engine's output really doubles your productivity for the type of quality that you yourself would be satisfied with. And you don't have to do the entire hour-long test either -- just do 20 minutes of it, and if you find that it's not helping, simply stop the test and tell the client "sorry, but your MT engine isn't good enough to justify 50%".
| | |
|
|
|
Samuel Murray
Netherlands
Local time: 18:15
Member (2006)
English to Afrikaans
+ ...
jyuan_us wrote: philgoddard wrote:
...that this text is for a skincare company. All the texts I've ever done for this industry are marketing copy, for which MT is wholly unsuitable. ... Machine translation is for getting the gist of a document. Human translation is for finely crafted prose that persuades people to buy things. The answer to that question is apparent.
Erm, sorry... what is the question? (I don't see a question in Phil's comment.)
| | | | Phil Hand
China
Local time: 00:15
Chinese to English
| Not just manufacturers | Oct 2, 2014 |
Tom in London wrote:
Do you find that Chinese manufacturers rely a lot on MT?
If you've ever seen those funny lists of bad translations with "f*** fruit" (because the Chinese word for dry is the same as the word for "do," and do can have the same sexual meaning it has in English), then you've seen a lot of Chinese MT. It's rampant.
| | | | | Google Translate is crowdsourcing with random computers | Oct 2, 2014 |
... and the results are just as random.
Or like setting chimps to translate Shakespeare. Chimps have come a long way since Washoe http://en.wikipedia.org/wiki/Washoe_(chimpanzee) and they ARE intelligent, but they are still not actually producing world literature in sign language.
Using a dedicated MT engine with controlled input and limited subject area, based... See more ... and the results are just as random.
Or like setting chimps to translate Shakespeare. Chimps have come a long way since Washoe http://en.wikipedia.org/wiki/Washoe_(chimpanzee) and they ARE intelligent, but they are still not actually producing world literature in sign language.
Using a dedicated MT engine with controlled input and limited subject area, based on human translation and CAT output, CAN give quite usable results. It requires a lot of human effort, and I am still surprised if it is actually viable, but for the 'here today, mulched tomorow' kind of stuff, it is still a lot better than Google. There are simply not enough humans with time and patience to fill the demand.
If an MT engine works largely on whole sentences, the grammar and syntax are OK in some pairs.The languages I work with - English and Danish - are not highly inflected, and I can imagine it would be a far greater problem just in French or German, quite apart from languages that are not closely related.
There is an enormous difference between output from Google Translate and the texts I was asked to look at, where there were a lot of standard expressions
-- road traffic regulations, international safety sentences and suchlike.
The language there may not be as elegant and idiomatic as you would like, but the fixed formulations have to be used, and a machine can match them up quite reliably.
BTW my client offered me more than 50% of the rate for translating from scratch. (I don't remember the percentage we finally agreed on, but it was more like 65.)
As I mentioned earlier, it is a completely different way of thinking from translating, and if it is not for you, then fine, go and do what you are good at.
You would not dream of knocking interpreters, just because you don't interpret yourself.
Knocking the results of Google and its ilk is fair enough, but we have to accept that MT is not going to go away. We need to educate clients about when it is not adequate and why. It is part of explaining why translation is not fast and easy, and why we need time and resources and to be paid for our work.
We also need to know in detail, and be able to convince clients in a world where many people trust their computer more than their neighbour, and websites more than their doctor.
Just dismissing all the serious work that has been done as totally useless is not going to convince people. They can see the results for themselves, and see the difference, so we need to look at it too.
As I said, I have changed my opinion over the years. I still don't think MT is going to put well qualified human translators out of work any time soon, but the younger generation at least will need to need to look at it and take it seriously. ▲ Collapse
| | | | | Pages in topic: < [1 2 3] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Response to Post-Machine Translation Editing Request at 50% Discount of Standard Rates | 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 
Translation news
