| Pages in topic: < [1 2 3 4] | post-editing of machine translations - Are there any real time savings? Thread poster: Thomas Johansson
|
|---|
Lieven Malaise
Belgium
Local time: 13:06
Member (2020)
French to Dutch
+ ...
[quote]Lieven Malaise wrote:
[Edited at 2022-05-23 06:39 GMT]
| | | | | You'll have to ask | May 23, 2022 |
Baran Keki wrote:
if translators, by using such tools, send their translated segments directly into Google's, DeepL's databases and thus, wittingly or unwittingly, contribute to the improvement of MT engines
You'll have to ask the developers of the CAT tool and/or the developers of the plugins that are used.
I've asked the developer of CafeTran Espresso and he assures me that no information from the Target Segment Editor is sent back to the MT providers.
I don't know what the others developers do.
| | | | Lieven Malaise
Belgium
Local time: 13:06
Member (2020)
French to Dutch
+ ...
| Fully agree. | May 23, 2022 |
expressisverbis wrote:
In short, the only real time saving when using post-editing is to have an excellent command of the particular area(s) you are translating, and know how to use Machine Translation Post-Editing in a technical and professional way.
That means to ensure the correct terminology, tone of voice, consistency throughout the text, and that it will meet the intended purpose and target audience.
All the rest is bullshit. That's it.
Amen.
| | | | | No reason for discounts | May 23, 2022 |
I've noticed that machine translations into Finnish are mostly complete rubbish, so I don't usually accept MTPE assignments anymore. However, I make an exception for one good client of mine, who pays a staggered rate for the post-editing. The more changes I make the higher the word rate becomes, and from 40% upwards they pay me my full word rate. I wouldn't commit myself to a fixed discount for MTPE, as I would just end up working as hard or usually even harder than if I was to translate the tex... See more I've noticed that machine translations into Finnish are mostly complete rubbish, so I don't usually accept MTPE assignments anymore. However, I make an exception for one good client of mine, who pays a staggered rate for the post-editing. The more changes I make the higher the word rate becomes, and from 40% upwards they pay me my full word rate. I wouldn't commit myself to a fixed discount for MTPE, as I would just end up working as hard or usually even harder than if I was to translate the text from scratch myself, and all for less money. ▲ Collapse
| | |
|
|
|
Lieven Malaise wrote:
It's perfectly possible to disagree on something without burying your discussion partner(s) into the ground. You should try it some time. As you can see above, the opposite is dead easy. Be civil.
Sorry, I shouldn't have been so obnoxious. My sincere apologies to everyone offended.
| | | |
Ice Scream wrote:
As I dictate, I reckon the time saving will be negligible.
Try dictating words like "jejunoileal", "vepoloxamer", "obinutuzumab" or "inguinodynia". See how much time you save.
Sometimes I have no idea how the English words are pronounced, especially in the case of new drug names. In order to use dictation, I would have to look the pronunciation up, which rather defeats the object. Not to mention training the software to understand a word I might never see again. Yet on the whole, MT handles that kind of terminology very well.
So my own "reckoning" has led to a different conclusion.
| | | | | MT vs PEMT vs HT - my version (HT) | May 25, 2022 |
Stepan Konev wrote: Thomas Johansson wrote:
It's an important issue, since the new machine translation technology is an important technology that probably will take over more and more of our industry - but if the time savings are an illusion, then there is no justification for applying or giving discounts. I have just machine-translated a paragraph from your message. Here it goes: Это важный вопрос, поскольку новая технология машинного перевода является важной технологией, которая, вероятно, будет захватывать все большую часть нашей отрасли, но если экономия времени является иллюзией, то нет никаких оснований для применения или предоставления скидок. It only requires minor edits. Even punctuation (which is very complicated in Russian) marks are used correctly here in this translation. For a professional translator it takes seconds to make it read well: Это важный вопрос, поскольку новые технологии машинного перевода — это серьезные технологии, которые, вероятно, будут захватывать все большую часть нашей отрасли. Но если экономия времени — иллюзия, то и никаких оснований для применения или предоставления скидок нет.[Edited at 2022-05-18 15:08 GMT]
Here's how I'd translate this one into my native Russian [peer feedback is welcome]:
EN: It's an important issue, since the new machine translation technology is an important technology that probably will take over more and more of our industry - but if the time savings are an illusion, then there is no justification for applying or giving discounts.
RU: Это важный вопрос, ведь значимость новой технологии автоперевода очевидна в контексте возможного увеличения ее веса в нашей отрасли. Однако если приписываемая ей экономия времени - всего лишь иллюзия, то и о правомерности применяемых и предоставляемых скидок говорить не приходится.
Three things I usually keep an eye on to stay MT-proof and optimistic about the future of translation:
1. Meaning: translate 'from the core' instead of messing with word salad.
2. Messaging: focus on pragmatics, not mechanics of the language.
3. Idiom: strive to be an authentic writer.
| | | | William Liu
Taiwan
Local time: 19:06
English to Chinese
+ ...
| The machine of no brain translates, you edit. | May 25, 2022 |
As a translator, I don't regard MTPE as a translation task. they are totally different jobs.
MTPE requires the Pro in specific area for the content or it takes much more time than the normal translation.
Based on the experience and background, you also need to:
Correcting syntax and semantic errors.
Adjusting the style and tone.
Ensuring all key terminologies are accurate.
Deleting any offensive or culturally inappropriate parts of the raw text.<... See more As a translator, I don't regard MTPE as a translation task. they are totally different jobs.
MTPE requires the Pro in specific area for the content or it takes much more time than the normal translation.
Based on the experience and background, you also need to:
Correcting syntax and semantic errors.
Adjusting the style and tone.
Ensuring all key terminologies are accurate.
Deleting any offensive or culturally inappropriate parts of the raw text.
Humanizing the raw MT output.
Moreover, the weighted rate based on their fuzzy match is extremely not worthy.
[修改时间: 2022-05-25 15:23 GMT] ▲ Collapse
| | |
|
|
|
Tony Keily
Local time: 13:06
Italian to English
+ ...
| Back in the day... | May 26, 2022 |
I remember doing EU institutional MTPE in Dublin in the 1990s. There was an understanding that the end product wouldn't be up to human translation standard. I guess if there's some kind of rider that things can be rough around the edges, that could be factored in, but it seems a sort of cack compromise and I never liked it.
DeepL isn't awful at EU institutional stuff, but bringing the output up to standard still takes about the same time as human translation.
| | | | Adieu 
Ukrainian to English
+ ...
I've seen some indications that TM entries MAY leak from MemoQ (or possibly from other productivity software or services clients use to mail, access, or manage files or translation memories).
Specifically, Google Translate has started sounding, capitalizing, and punctuating a lot like my edits on some niche boilerplate text. Which is pretty odd since I generally work with internal documents that never get published anywhere.
Then again, I have no idea if you'd actually ... See more I've seen some indications that TM entries MAY leak from MemoQ (or possibly from other productivity software or services clients use to mail, access, or manage files or translation memories).
Specifically, Google Translate has started sounding, capitalizing, and punctuating a lot like my edits on some niche boilerplate text. Which is pretty odd since I generally work with internal documents that never get published anywhere.
Then again, I have no idea if you'd actually get the same output if you plugged the same text into Google Translate on another IP logged into a different account. Would you?
Hans Lenting wrote: Baran Keki wrote:
if translators, by using such tools, send their translated segments directly into Google's, DeepL's databases and thus, wittingly or unwittingly, contribute to the improvement of MT engines You'll have to ask the developers of the CAT tool and/or the developers of the plugins that are used. I've asked the developer of CafeTran Espresso and he assures me that no information from the Target Segment Editor is sent back to the MT providers. I don't know what the others developers do. ▲ Collapse
| | | | | Not a tool, but a product | Aug 13, 2022 |
Stepan Konev wrote:
If you can use the tool, it works for you well. If you can't, it doesn't.
Many translators complained that CAT tools drastically slow them down when their employers forced them to use such tools. Nowadays, they can't live without their favorite CAT tools. The same applies to MT. If you are not trained enough to use it, it will be useless to you. Actually it works for any tool in general, not only those for translation.
Let's not mix up different things. The topic starter's concern was about time saving and rate reduction when using a machine translation offered for post-editing by a translation agency (or an end client). In this event it is not a tool, but a product, like premade dough or pastry, from which the cook (the translator in our case) is supposed to bake a nice, tasty cake (perfect translation). It is machine-made dough, which may be stiff or wet, hard or soft, yeasted or unyeasted, biscuit or shortcrust, etc. In short, it may be anything between very suitable for baking the final product (a cake) or quite unsuitable for it. Hence the labor and time inputs by the cook/translator to achieve the desired result may vary greatly.
If the client uses a generic machine translation engine like Google Translate, or DeepL, or Yandex to prepare the text for post-editing, why should translator agree to or offer any discounts? He or she may just as well use such an engine on his/her own, spending just a few minutes for several pages. In this case the use of machine translation forms an integral part of translator's internal work process and is indeed a tool. But the translator's internal work process should be of no concern to the client, who should be interested only in the quality of the final product. Why give discounts then?
It's another pair of shoes when the client (usually a translation agency) uses a custom MT engine, well-trained for specific content, to produce a nearly seamless translation, which needs only some final polishing. In this case a discount may be a subject for discussion. But I've personally never seen such translations as yet.
[Edited at 2022-08-13 11:41 GMT]
[Edited at 2022-08-13 11:43 GMT]
| | | |
Vladimir Morozov wrote:
If the client uses a generic machine translation engine like Google Translate, or DeepL, or Yandex to prepare the text for post-editing, why should translator agree to or offer any discounts? He or she may just as well use such an engine on his/her own,
…
It's another pair of shoes when the client (usually a translation agency) uses a custom MT engine, well-trained for specific content, to produce a nearly seamless translation, which needs only some final polishing. In this case a discount may be a subject for discussion. But I've personally never seen such translations as yet.
Neither have I.
And regarding to the use of MT by freelance translators: it is better not to pretranslate the whole text with one engine. Instead, move from segment to segment, and let several engines make suggestions. So texts that are pretranslated by an agency are counterproductive and require extra effort, and thus you should charge extra.
| | |
|
|
|
| | Marta Kiezik
Poland
Local time: 13:06
English to Polish
+ ...
| No, not a lot of savings really | May 5, 2023 |
My problems with NMT/CATs + technical texts are the following:
a) bad results if the source text was not written by a native speaker: a fresh translation would be a much better idea in this case;
b) a lot of time lost/unpaid if a project consists of a gazillion of short files (say, a technical documentation), and terminology inconsistencies across the project files make it even worse;
c) agencies paying 20% for fuzzy matches (which, for reasons unknown to me, often conta... See more My problems with NMT/CATs + technical texts are the following:
a) bad results if the source text was not written by a native speaker: a fresh translation would be a much better idea in this case;
b) a lot of time lost/unpaid if a project consists of a gazillion of short files (say, a technical documentation), and terminology inconsistencies across the project files make it even worse;
c) agencies paying 20% for fuzzy matches (which, for reasons unknown to me, often contain erroneous data absorbed from the original fuzzy segment, making for critical errors that can cost you your honorarium, plus by correcting these, you are doing them an additional, unpaid service);
d) agencies not paying anything at all for a 100% match/context match: in my experience, these happen often in segments that have been hastily pre-approved by the PMs, especially in the case of generic headlines with numerous possible meanings;
e) ENORMOUS cognitive effort that proofreading a machine-translated text containing non-intuitive, twisted errors requires, and it also takes WAY more time than standard proofreading;
f) hallucinations and wrong terminology as AI never leaves a blank space; thus, more likelihood of a critical error evading the attention of both the posteditor and the proofreader, which means increased likelihood of legal risks to the agency;
g) natural translation flow and job satisfaction are axed.
I've been recently offered to lower my NMT discount from 30% to 40%, with the agency citing 'better engine efficacy'. Needless to say and given the current raging inflation rates, I rejected and as of now, haven't heard from them yet. I think this might turn into a quality gap between agencies and freelancers in the future. Might not be that bad for the latter. ▲ Collapse
| | | | | 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 » post-editing of machine translations - Are there any real time savings? | 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 » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
