| Pages in topic: [1 2] > | Poll: Do you accept projects that involve post-editing machine translated content? Thread poster: ProZ.com Staff
|
|---|
This forum topic is for the discussion of the poll question "Do you accept projects that involve post-editing machine translated content?".
This poll was originally submitted by Montse S.. View the poll results »
| | | | neilmac
Spain
Local time: 12:28
Spanish to English
+ ...
| Not if I can help it | Feb 12, 2015 |
In principle, no, because I see that as akin to cutting one's own throat if you're a translator. However, most of the text revision work I do is from direct clients in academia who write their own papers in English, and I assume they will use all the resources available to them, including MT applications such as Google translate. In fact, there are certain errors which I usually ascribe to MT, but I don't make a fuss about it.
However, if the query refers to the plethora of agencies and ot... See more In principle, no, because I see that as akin to cutting one's own throat if you're a translator. However, most of the text revision work I do is from direct clients in academia who write their own papers in English, and I assume they will use all the resources available to them, including MT applications such as Google translate. In fact, there are certain errors which I usually ascribe to MT, but I don't make a fuss about it.
However, if the query refers to the plethora of agencies and others currently foisting MT texts onto desperate translators/reviewers, then nay, nay and thrice nay, sirrah. ▲ Collapse
| | | | | What's the point? | Feb 12, 2015 |
There is no point in post-editing machine-translated content, because either you have the source language text and therefore you need to translate it and then write your translation in place of the machine-generated one, or you don't have it and you need to guess what the source text was in any case. Both of these approaches usually require more work than translating the text directly, and lead to worse results.
| | | | | I don't work for any customers that work in this way, but | Feb 12, 2015 |
I think they'd know better than to ask me anyway
| | |
|
|
|
I've been postediting machine translation for 35 years. It's not my favorite thing to do, but the system my client uses is excellent and sometimes the output requires no editing. It can speed me up or slow me down, depending on the input text and the "machine's" performance. I would prefer to translate from scratch. My human translations are more organic and expressive.
Let's put it this way: My first choice is human translation, but I would rather postedit the output of a *good* MT... See more I've been postediting machine translation for 35 years. It's not my favorite thing to do, but the system my client uses is excellent and sometimes the output requires no editing. It can speed me up or slow me down, depending on the input text and the "machine's" performance. I would prefer to translate from scratch. My human translations are more organic and expressive.
Let's put it this way: My first choice is human translation, but I would rather postedit the output of a *good* MT system than format and translate a PDF. I would rather postedit a document of running text than hand-translate an academic transcript or a legal certificate. It's all relative.
[Edited at 2015-02-12 09:19 GMT] ▲ Collapse
| | | | | I can see your point | Feb 12, 2015 |
Muriel Vasconcellos wrote:
I've been postediting machine translation for 35 years. It's not my favorite thing to do, but the system my client uses is excellent and sometimes the output requires no editing. It can speed me up or slow me down, depending on the input text and the "machine's" performance. I would prefer to translate from scratch. My human translations are more organic and expressive.
Let's put it this way: My first choice is human translation, but I would rather postedit the output of a *good* MT system than format and translate a PDF. I would rather postedit a document of running text than hand-translate an academic transcript or a legal certificate. It's all relative.
[Edited at 2015-02-12 09:19 GMT]
Thanks for being honest, Muriel
| | | |
I have looked at it a few times - either because there is no getting round it, and I can imagine there are highly repetitive texts where it might work. Otherwise, it is always a good strategy to know your enemy. I normally avoid it and here are some reasons:
It has numerous weaknesses - trouble with homonyms is one. The results can be from hilarious til disastrous.
It cannot 'read between the lines' and translates literally. Same comment as above.
It skips ... See more I have looked at it a few times - either because there is no getting round it, and I can imagine there are highly repetitive texts where it might work. Otherwise, it is always a good strategy to know your enemy. I normally avoid it and here are some reasons:
It has numerous weaknesses - trouble with homonyms is one. The results can be from hilarious til disastrous.
It cannot 'read between the lines' and translates literally. Same comment as above.
It skips terms it doesn't recognise - and of course, those include all the difficult ones, though if you know the subject area, it may be easy enough to fill some in.
The logic is not human, so it is difficult to find rhyme or reason in the mistakes it makes.
One phrase about Swedish executives wearing shorts for security reasons had me puzzled until I managed to get hold of the source. In Stockholm, people just don't attend top level meetings in shorts... No, they were carrying (ID) cards.
Etc.
If MT uses a CAT principle with whole segments, and IF the source is pre-edited to use controlled language and avoid known problems, you can get some almost usable results.
You can aslo get some text that looks fine, but is nothing like an accurate translation of the source, and it really is HARD WORK checking it. People have measured and counted their way to the conclusion that it is no worse than the lower levels of human translation... So in theory you can get used to post-editing it, and in blind tests, people asked to distinguish human translation from MT were wrong around 50% of the time!
Having seen some human translations that were utter garbage, I wonder why anyone would use machines to produce even more of it??
At the present state of the game, it is easier to get the humans to improve than the machines - I have seen it happen over the years.
Language works by human logic, or possibly no logic at all. 
Some of the best jokes are double entendre or when the logic goes awry, but they are often untranslatable.
Language just does not fit into algorithms and binary logic.
[Edited at 2015-02-12 11:51 GMT] ▲ Collapse
| | | | | Not so far... | Feb 12, 2015 |
I have never been asked, but I doubt that I would as I refuse most proofreading projects...
| | |
|
|
|
| Once in a while... | Feb 12, 2015 |
...if they approve my rate, because I charge full rate and translate it again.
| | | | Noura Tawil 
Syria
Local time: 13:28
Member (2013)
English to Arabic
The question is not even applicable in my language pair.
Machine translations into Arabic are too poor to be of any good to a post-editor. Period!
| | | |
I haven't been offered yet; I have looked into some projects that were on some platforms, but the remuneration is way too low. If that is adjusted, I might give it a try in the future, but I am not not holding my breath
| | | | DianeGM
Local time: 13:28
Member (2006)
Dutch to English
+ ...
| Not consciously ... | Feb 12, 2015 |
In the past I had one client that I that for. The projects concerned medical questionnaries which were used for analysis purposes. The purpose of the post-editing was basically to identify and fix the completely nonsensical parts - not to get the entire machine output text to the level of a human translation.
I've never been offered any other job explictly stating that it was post-editing machine translation - though I think I've also unknowingly taken on exactly that task under the... See more In the past I had one client that I that for. The projects concerned medical questionnaries which were used for analysis purposes. The purpose of the post-editing was basically to identify and fix the completely nonsensical parts - not to get the entire machine output text to the level of a human translation.
I've never been offered any other job explictly stating that it was post-editing machine translation - though I think I've also unknowingly taken on exactly that task under the guise of 'proofreading'.
Nowadays I rarely take on any proofreading or editing jobs whatsoever.
[Edited at 2015-02-12 12:05 GMT] ▲ Collapse
| | |
|
|
|
Muriel Vasconcellos wrote:
I've been postediting machine translation for 35 years. It's not my favorite thing to do, but the system my client uses is excellent and sometimes the output requires no editing. It can speed me up or slow me down, depending on the input text and the "machine's" performance. I would prefer to translate from scratch. My human translations are more organic and expressive.
Let's put it this way: My first choice is human translation, but I would rather postedit the output of a *good* MT system than format and translate a PDF. I would rather postedit a document of running text than hand-translate an academic transcript or a legal certificate. It's all relative.
[Edited at 2015-02-12 09:19 GMT]
Same here, although I've only done this occasionally.
[Edited at 2015-02-12 14:56 GMT]
| | | | Teresa Reinhardt
United States
Local time: 03:28
Member (2002)
German to English
+ ...
[quote]neilmac wrote:
In principle, no, because I see that as akin to cutting one's own throat if you're a translator.
I can only agree. I used to have to do it because my first FT employer had its own
system developed…only to have to shut it down soon because German and English syntax function very differently. It took longer to fix the stuff (than anticipated - for the supposed "leverage" they had been promised).
Now, I don't do it at all. Linguistics departments have been working on this for a long time, and it may work for very sub-standard expectations and on very simple/simplified texts. But it's hard to get those from writers consistently. And it's not something you
would want to show anyone in any language. Communications are part of a company's image, and shortcuts don't usually pay in that area.
Now, corpus-based systems are a different thing, but fortunately Google got distracted by the next shiny thing. Or else we would already have a very competitive system.
| | | |
to do so if it involves languages other than Romance languages, and even then they'd be more trouble than they're worth. MT is basically useless in technical subjects which is where most of us earn our living. Have you ever seen machine-translated engineering specs? Only the inexperienced, destitute and/or naive would undertake it...
| | | | | Pages in topic: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: Do you accept projects that involve post-editing machine translated content? | Trados Studio 2022 Freelance | The leading translation software used by over 270,000 translators.
Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop
and cloud solution, empowering you to work in the most efficient and cost-effective way.
More info » |
| | CafeTran Espresso | You've never met a CAT tool this clever!
Translate faster & easier, using a sophisticated CAT tool built by a translator / developer.
Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools.
Download and start using CafeTran Espresso -- for free
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
