| Pages in topic: < [1 2 3 4] > | Has using DeepL become the norm? How to calculate PEs? Thread poster: dieuebersetzung
|
|---|
AnnaSCHTR 
United States
Local time: 21:56
English to Czech
+ ...
| Not a norm for many of us | Jul 16, 2021 |
Due to some clients’ privacy concerns, we never use any MT unless specifically asked to do so (which happens very rarely).
To say that it doesn’t matter what tool is used seems somewhat odd to me. Many of our projects are of a highly confidential nature and the clients expect that we handle them in a certain manner.
Whether or not their privacy concerns are justified it is not up to me to say.
I would recommend that you add a specific language to your purchase orders rega... See more Due to some clients’ privacy concerns, we never use any MT unless specifically asked to do so (which happens very rarely).
To say that it doesn’t matter what tool is used seems somewhat odd to me. Many of our projects are of a highly confidential nature and the clients expect that we handle them in a certain manner.
Whether or not their privacy concerns are justified it is not up to me to say.
I would recommend that you add a specific language to your purchase orders regarding the use of MT, or discuss it with the translators in advance. ▲ Collapse
| | | | | Can still be handled with DeepL | Jul 16, 2021 |
AnnaSCHTR wrote: Many of our projects are of a highly confidential nature and the clients expect that we handle them in a certain manner.
... unless you use the DeepL API, as an interface between DeepL and your CAT. But there are (still) other methods to use DeepL, even for free, and I don't mean the copy-and-paste window on their website.
| | | | AnnaSCHTR
United States
Local time: 21:56
English to Czech
+ ...
| I agree Matthias but... | Jul 16, 2021 |
Yes, absolutely, there are ways to work safely with API (and probably by other methods as well). I think the MT development is fascinating and will radically change we all work in a foreseeable future.
However, I feel that for now we need to respect clients' feelings regarding these matters.
We are bombarded daily by catastrophic news about security breaches, data leaks etc. The progress of our thinking is a bit slower than the development of technology.
... See more Yes, absolutely, there are ways to work safely with API (and probably by other methods as well). I think the MT development is fascinating and will radically change we all work in a foreseeable future.
However, I feel that for now we need to respect clients' feelings regarding these matters.
We are bombarded daily by catastrophic news about security breaches, data leaks etc. The progress of our thinking is a bit slower than the development of technology.
Just like with many other things, it is better to be upfront and make sure there is a mutual agreement between those who do the work and those who pay their fees. ▲ Collapse
| | | | Stepan Konev
Russian Federation
Local time: 05:56
English to Russian
| Confidentiality vs similarity | Jul 16, 2021 |
AnnaSCHTR wrote:
we never use any MT unless specifically asked to do so When my clients ask me not to use MT, I don't. But why not use it when they don't ask? I have an NMT engine (OPUS CAT MT Engine) that doesn't even need Internet connection. It is fine-tuned using my TM and generates translations based on my TM. Also, there are other measures to keep data secured when using online MT engines.
However, as far as I understand the original post, we deviate again toward confidentiality. The major concern is that the topic starter's clients can compare DeepL output with final translation and turn upset because they see similar words and letters there... Well, ok, even if DeepL output looks similar to the final translation, so what? If the translation is good, the translator's work is done. Period.
Another concern expressed above was that a translator completes his/her work much faster and that is why it is not fair to pay him/her the same price as a translator who completes his/her work much longer. This standpoint is just beneath any criticism.
[Edited at 2021-07-16 16:36 GMT]
| | |
|
|
|
| Lots of similes | Jul 16, 2021 |
Ok, I guess it's difficult to discuss this objectively in a translator forum
I am not sure if you have considered this from the perspective of the end customer who knows that MT with PE takes a lot, and I mean a lot less time and ask himself why it is ok that he continues to pay the same price.
Of course there are texts that are not suited well for it but the things that I have seen were very, very close to the MT and t... See more Ok, I guess it's difficult to discuss this objectively in a translator forum
I am not sure if you have considered this from the perspective of the end customer who knows that MT with PE takes a lot, and I mean a lot less time and ask himself why it is ok that he continues to pay the same price.
Of course there are texts that are not suited well for it but the things that I have seen were very, very close to the MT and these people weren't even using licenses but the free copy-and-paste version. For that, I am not paying (almost) what I pay for a human translation. And frankly I don't see this could be a viable business model. ▲ Collapse
| | | | Abba Storgen (X)
United States
Local time: 21:56
Greek to English
+ ...
| "Where is this industry heading?" | Jul 16, 2021 |
dieuebersetzung wrote:
And out of five, all used DeepL without telling me before and while charging rates that I associate with human translation (8 to 12 cents per word).
Correct. I see it in 100% of jobs I edit. Sentences that are either too long (unnecessary words, boring "teleprompter" type writing, errors from previous translators using Google constantly there, and they all sound the same and you can't concentrate because the text has no contrast and flow). A sentence that should be "download the file and store it safely" reaches my desk as "make downloading of the file and warehouse it in a safe way". Now, imagine 500 sentences like that, in a row. I've given up, I've damaged my health enough. Who's going to read it anyway... if bubblegum Taloula is buying memberships and ribbons left and right, writes garbage with a shiny smile and gets recommended from like-minded peers, I can't fight that.
This industry will have the same fate as pop music had in the last 50 years. From many independent talented middle-class creators, to just a handful of large companies producing machine content. And none of talented creators in sight. Only shallow marketing.
If I could turn time back 25 years, I would gladly accept to repeat all other life mistakes, except this one: my entry into this industry.
PS. They all lie, they're all using it. I've made myself an expert in detecting the tell-tale signs, and they're plenty.
| | | |
If you ask for a translation, you'll likely get an MT-assisted translation. The piece can be very good or very bad.
If you ask for transcreation, you're likely to pay more, and less likely to see MT through it. The piece can be very good or very bad.
I wouldn't want to be an agency.
I am part of those who don't see any benefits in displaying MT suggestions in CAT tools or pretranslating with MT. I use Dragon for speed and convenience, and I like to keep complete freedom ... See more If you ask for a translation, you'll likely get an MT-assisted translation. The piece can be very good or very bad.
If you ask for transcreation, you're likely to pay more, and less likely to see MT through it. The piece can be very good or very bad.
I wouldn't want to be an agency.
I am part of those who don't see any benefits in displaying MT suggestions in CAT tools or pretranslating with MT. I use Dragon for speed and convenience, and I like to keep complete freedom in how I convey a piece. With MT suggestions, I fear I'd be more lenient.
However, I am aware that some colleagues can make skilled use of MT suggestions. We're certainly entitled to use whatever tools that speed up the job. Some can't work efficiently with speech recognition or CAT tools, others can't with MT suggestions.
I haven't tried off-line self-standing local MT engines trained with my TMs, but more because I don't have time to look into that (too much human translation work in my pair!) than ideological lock. And if I used it, it certainly wouldn't warrant any lower price for my service.
In EN>FR editing or on-boarding assignments I get, exclusively from agencies with high standards and loyal to their translators, I do smell MT more often than say 10 years ago. The outcome is thick, doesn't flow and can be downright misunderstood. It often goes hand in hand with average source language understanding.
dieuebersetzung wrote:
....One French translator ... charged 12 cents per word and delivered a translation that was so clearly MT that he not even bothered to deny it.
Most juggernauts offer 8, 9, 10 max after serious negotiation for flowing EN>FR human translations (customer-facing stuff).
Agencies accepting 12 for EN>FR typically focus on quality, long-term partnerships and should be pampered instead. Some people never understand how to do business.
dieuebersetzung wrote:
...It's about having a 50 page document that would take 7 days to translate using human translation vs. no time with MT and then maybe a day's work for PE...
From my MTPE experience 10-15 years ago (agency assignments), you save 50% of your time using purpose-made MT, not the readily available options like DeepL, and deliver "good enough" stuff. The more thorough (read: customer-oriented) the translator, the less time they save.
Philippe ▲ Collapse
| | | | Samuel Murray
Netherlands
Local time: 04:56
Member (2006)
English to Afrikaans
+ ...
dieuebersetzung wrote:
And out of five, all used DeepL without telling me before and while charging rates that I associate with human translation (8 to 12 cents per word).
Strictly and theoretically speaking, you're paying for the end-product, so it should not matter if a translator uses a machine-translator to produce a human-quality translation.
However, in reality, many translators misjudge how close to human-quality their post-edited machine-translated end-products actually are. DeepL can create a very human-like translation on a per-sentence basis, but (as you and others have also mentioned) consistency of terminology or consistency of phrasing is something that the translator has to ensure, and the less consistent DeepL's translations, the more work the translator needs to put in to make the text better.
I know of many translators who use machine-translation these days, and many translators use it without telling the client and even if the client asked them not to. I don't know what is the solution to this problem.
[Edited at 2021-07-16 18:17 GMT]
| | |
|
|
|
...to the new trend of PEMT and change your brand name by including an underscore (die_uebersetzung), so that it would sound in English like "Die, translation!" (Stirb, Übersetzung!). And as an agency, you could go even further (die_uebersetzer).
| | | |
Samuel Murray wrote:
However, in reality, many translators misjudge how close to human-quality their post-edited machine-translated end-products actually are. DeepL can create a very human-like translation on a per-sentence basis, but (as you and others have also mentioned) consistency of terminology or consistency of phrasing is something that the translator has to ensure, and the less consistent DeepL's translations, the more work the translator needs to put in to make the text better.
DeepL can save you key strokes, just like auto-completion and auto-assembling can. But you’ll still have to put everything together. And make the target text consistent and a good or better representation of the source.
The time saving isn’t as huge as Robert may assume. I’d say it’s more like 10 % (compared to AC and AA).
| | | | Edward Potter
Spain
Local time: 04:56
Member (2003)
Spanish to English
+ ...
| Another anecdote | Jul 17, 2021 |
A relative of mine once told me of a Cousin who wanted him to install a phone for her back in the 1990s. My relative was a specialist on that field and he knew what to do. When he was ready, the Cousin said: "Oh! Had I known that it was so simple, I could have done it myself!" The Point is, she didn't know. And that knowledge is what has to be paid for.
A friend of a distant relative of mine was a mechanic. A customer came in one day saying his car was making a strange noise. The mechanic goes over to the car, taps a part a few times, comes back over and says, "that will be 500 dollars". The customer says, "but you just tapped something with a hammer a few times. Why is it so expensive?" The mechanic replied, "yes, but I knew where to tap."
| | | | | Additional questions | Jul 17, 2021 |
Hi Robert,
In your OP, you mention five cases of new collaborations where MT (DeepL) was systematically used for the translation, with only minimal (if at all) editing.
In one of them, the French translator delivered mostly unchanged MT output (around 90%), where "even the PE was done sloppily, with inconsistent terminology that was easy to spot".
What about the rest? As far as you and your internal team can estimate, was the content fit for purpose? Was it... See more Hi Robert,
In your OP, you mention five cases of new collaborations where MT (DeepL) was systematically used for the translation, with only minimal (if at all) editing.
In one of them, the French translator delivered mostly unchanged MT output (around 90%), where "even the PE was done sloppily, with inconsistent terminology that was easy to spot".
What about the rest? As far as you and your internal team can estimate, was the content fit for purpose? Was it of deliverable, professional, "human" quality?
For now, all we know for sure is about the French case, for which we can safely assume that the end product was subpar.
The translator took a shortcut, but the result was not there. So how much time or effort it took is irrelevant.
Has any of these translators (and how many, out of your five sample case study) delivered quality work as was expected for the price paid?
With this information, maybe more useful insights can be gained from this discussion.
[Edited at 2021-07-17 11:51 GMT] ▲ Collapse
| | |
|
|
|
| Incorrect assumptions | Jul 17, 2021 |
dieuebersetzung wrote:
I am not sure if you have considered this from the perspective of the end customer who knows that MT with PE takes a lot, and I mean a lot less time and ask himself why it is ok that he continues to pay the same price.
This is a misconception which is unfortunately very widespread and causing a lot of distortion in the market.
Please note: MTPE only saves time (and money), compared to original human translation, if you accept lower quality and the occasional mistranslation.
Print this sentence out and hang it over your monitor. Add it to your email signature and send it as a christmas card to all your clients (don't wait until Christmas, too).
Human translation means that a process takes place in the mind of the human translator, consisting of understanding the source text and reproducing it in the target language, taking into account things like purpose and intentions, target audience, subject matter knowledge, cultural differences, glossaries, style guides and other client instructions, and a lot more. This is the main work the human translator does, not writing down the result.
If you add MT, that does not change a thing. The translator still has to do all these thought processes above, MT does not save anything of it. On the contrary: you are adding more work since you ask the translator to compare their own result with the MT result. The only way to actually reduce the workload of the translator is to expressly ask them not to do their original translation process described above, and simply proofread the MT result instead, ignoring the source text except if the MT output makes no sense. This would be the only chance to save time and effort using MT. And of course you sacrifice a lot of quality and allow for mistranslations in this case. This can be done but it is simply a different quality level, and everyone involved should be aware of that. If you want the quality of human translation, MTPE does not save time and effort.
Now, it is still possible that the competent human translator and the unthinking MT arrive at the same result in the end. This might even happen with many sentences. But that does not mean anything. MT output is like a nice facade, it can look like the real thing, but some of these facades will simply topple over if you so much as touch them.
| | | | Jose Ruivo
Portugal
Local time: 03:56
Member (2007)
English to Portuguese
+ ...
| Once upon a time... | Jul 17, 2021 |
dieuebersetzung wrote:
I guess it depends a lot on what they ask for it too.
As an example, I also noticed two translators I've worked with for around 8 years are now using DeepL.
You say it's fair game, but I think that it's not ok to charge the same rate for a work that now only takes a fraction of the time.
Myself and many other translators I work with have not been able to update our rates for about 10 years. MT is very welcome and fair update to our income.
Once upon a time, I had clients who forbade using any kind of CAT tool, out of fear for lower quality translations, while I was already using it on demand from most clients. Is that an issue now? It will hapen the same with MT and MTPE.
It seems to me that your issue is more about the the money than the translation quality (or lack of it, of course). There probably will come a time when MT will come into the price equation, but I refuse to pass all the gains - 50% you say - from MT to the client only - we need at least to split the gain. Assuming your 50% gain calculation, 25% discount for you, and 25% gain for me.
| | | | Nadja Balogh
Germany
Local time: 04:56
Member (2007)
Japanese to German
+ ...
Even though it is true that there’s still a lot of effort that, at least theoretically, should go into a translation even while using MT, I do find it important to consider the end client too. I don’t know how many of them are aware of DeepL and such, but if they are, then of course they might – just like Robert – feel cheated if they find that their expensive translation is very similar to what DeepL is giving them.
It is always difficult to make comparisons with other fiel... See more Even though it is true that there’s still a lot of effort that, at least theoretically, should go into a translation even while using MT, I do find it important to consider the end client too. I don’t know how many of them are aware of DeepL and such, but if they are, then of course they might – just like Robert – feel cheated if they find that their expensive translation is very similar to what DeepL is giving them.
It is always difficult to make comparisons with other fields (the usual lawyers, dentists and plumbers) so please don’t be too strict here, but as a loose analogy, it’s like the difference between a bad restaurant and a good one – the bad one will use ready-made sauces and what do I know and will make a handsome profit out of it, and unless they claim not to use such ingredients that’s perfectly legal and I as a guest can’t complain. But I would still feel disappointed and would not return to that particular place.
So, returning to the translation field, maybe some clients won’t come back to the bland MT-“inspired” translation either, but just use DeepL themselves (if that’s what they’re going to get anyway). And this, in the long run, would really hurt us translators as well. ▲ 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 » Has using DeepL become the norm? How to calculate PEs? | 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 » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
