| Pages in topic: < [1 2 3 4 5 6 7] > |
Who is not using MT today? Thread poster: Gary Evans
|
|---|
Gary Evans 
Germany
Local time: 10:24
German to English
TOPIC STARTER | Confidentiality | Aug 13, 2019 |
Hi Dan,
I appreciate concerns about confidentiality, but (naming no names) there are MT engines which explicitly do not store any data as part of their offer. I've compared numerous comparisons of MTs in my research work and found a wide range of output quality. Of course the language combination is a key issue. I'm lucky. German to English is a widely translated combination.
Kind regards,
Gary
Dan Lucas wrote: Gary Evans wrote:
I've been testing machine translation tools for a while now. Considering how many CAT tools now have MT APIs, isn't the use of MT simply logical? So my question is: who is not using MT and why? I am not dead set against the use of MT, but the limited tests I have conducted in my pair and field of specialization suggest that it isn't that useful, as it required me to perform a lot of editing. It didn't feel to me as if I was getting any improvement in productivity. A year or so back, I was talking to the person who was heading up the internal MT effort at a very large Tokyo agency, and they told me straight out that their results had been poor, and not remotely ready for deployment. For gisting MT may be useful, but for my pair it doesn't seem suitable for output that is bound for a formal publishing process. Maybe this will change. More importantly, my clients are concerned about the misuse of (often confidential) data by MT services to the point that the majority of them explicitly forbid its use. Regards, Dan [Edited at 2019-08-13 16:26 GMT]
| | | |
Liviu-Lee Roth
United States
Local time: 04:24
Romanian to English
+ ...
| Absolutely not, for certain niches | Aug 13, 2019 |
As the Misha pointed out, there are certain fields where MT is not useful nor welcomed. I could name a few: translating for govt. agencies (DoD, DOJ, DHS, working docs for the FBI, USSS), other law enforcement agencies, courts, Department of State etc. One of their prerequisite is not to use CAT either.
Lee
| | | |
Gary Evans
Germany
Local time: 10:24
German to English
TOPIC STARTER | But if the MT does not store data... | Aug 14, 2019 |
Hi Lee,
This is not an issue if the MT does not store the data. This issue seems to be the only reason MT is not officially used. Wanna bet how many of these documents are being translated using MT without the knowledge of the client?
Looks like the US criminal justice system is keeping many people in work!
Liviu-Lee Roth wrote:
As the Misha pointed out, there are certain fields where MT is not useful nor welcomed. I could name a few: translating for govt. agencies (DoD, DOJ, DHS, working docs for the FBI, USSS), other law enforcement agencies, courts, Department of State etc. One of their prerequisite is not to use CAT either.
Lee
| | | |
Jan Truper
Germany
Local time: 10:24
Member (2016)
English to German
Gary Evans wrote:
So my question is: who is not using MT and why?
I am not using MT today.
I am doing subtitle translation today, and MT is utterly useless for this kind of work.
For the job at hand, the client actually offers a pre-translated template -- which means a little extra effort for me, since I have to trash it first before I can actually start working.
I was using MT yesterday.
I was doing game translation yesterday, and I pretranslated all segments with MT. I ended up overwriting the vast majority of them, because it takes less time to do so than to fix several words per sentence and/or to rearrange the sentence structure. If I had an intern sitting next to me making suggestions with that level of imbecility, I would send him home after three sentences: "Dude, sorry, this is just not working out."
MT may be useful for straightforward, simple or formulaic text -- but is has no wit, no taste, no style, no feel, no sensibility, no finesse, no sense of beauty or ugliness, is not aware of context, has no clue when to use formal/informal/singular/plural addressing, does not consider length or reading speed restrictions, is not able to split sentences over several segments, etc., etc.
MT is effing stupid.
I found The Misha's jackhammer/jeweller analogy quite fitting. I appreciate having a decent jackhammer at my disposal, but it only helps me to a very limited extent.
| | |
|
|
|
Gary Evans
Germany
Local time: 10:24
German to English
TOPIC STARTER | Meanwhile, MT is gaining ground | Aug 14, 2019 |
Hi Jan,
MT is a highly useful tool and it's transforming the industry. Read this:
https://www.emarketer.com/content/how-ai-and-machine-translation-are-changing-the-language-services-industry
It's not just MT which is gaining ground. Natural language processing (NLP) is also improving rapidly... See more Hi Jan,
MT is a highly useful tool and it's transforming the industry. Read this:
https://www.emarketer.com/content/how-ai-and-machine-translation-are-changing-the-language-services-industry
It's not just MT which is gaining ground. Natural language processing (NLP) is also improving rapidly. Those who adopt and make use of these tools will be around for some time to come as humans are still needed for "that human touch".
BTW which MT engines have you tried? Some really are crap while others are frighteningly good.
Regards from Aachen,
Gary
Jan Truper wrote: Gary Evans wrote:
So my question is: who is not using MT and why?
I am not using MT today. I am doing subtitle translation today, and MT is utterly useless for this kind of work. For the job at hand, the client actually offers a pre-translated template -- which means a little extra effort for me, since I have to trash it first before I can actually start working. I was using MT yesterday. I was doing game translation yesterday, and I pretranslated all segments with MT. I ended up overwriting the vast majority of them, because it takes less time to do so than to fix several words per sentence and/or to rearrange the sentence structure. If I had an intern sitting next to me making suggestions with that level of imbecility, I would send him home after three sentences: "Dude, sorry, this is just not working out." MT may be useful for straightforward, simple or formulaic text -- but is has no wit, no taste, no style, no feel, no sensibility, no finesse, no sense of beauty or ugliness, is not aware of context, has no clue when to use formal/informal/singular/plural addressing, does not consider length or reading speed restrictions, is not able to split sentences over several segments, etc., etc. MT is effing stupid. I found The Misha's jackhammer/jeweller analogy quite fitting. I appreciate having a decent jackhammer at my disposal, but it only helps me to a very limited extent. ▲ Collapse
| | | |
B D Finch
France
Local time: 10:24
French to English
+ ...
| Is farming an indicator of the future of translation? | Aug 14, 2019 |
Fatih Serdar ÇITAK wrote: I would like give an example. Did the sector of agriculture end when the machinery was introduced in it? No, it did not. Instead, this development oriented amateur agriculturers into new job fields, opened new business areas and only people that could cope with machinery knowledge became prominent in agriculture. I do think that this will be the case in our industry as well, not exactly the same of course but similar. Thank you for reading  Fatih Serdar Citak - An enthusiastic and eager translator.
Really? Agricultural machinery could have been used to improve the lives of farmers and agricultural workers. However, that would have depended upon it being introduced in a political economy focussed upon improving the lives of people in general. In fact, agricultural machinery has tended to reduce employment in farming, drive small farmers out of business or reduce their incomes, while providing large profits for multimational agribusiness. Translators are not likely to be the people who profit from MT under our current political economy. Mechanisation has also led to factory farming of livestock and mega farms. While farming has always involved cruelty to animals, the scale and intensification of that cruelty has become extreme, alongside the draining off of profits by multinational agribusiness. Personally, I don't fancy life as a battery hen. 
https://fullfact.org/economy/farming-subsidies-uk/
“Last year the average farm made £2,100 from agriculture and £28,300 from subsidies. The typical cereal farmer actually lost £9,500 by farming cereals.”
The Times, 4 August 2016
https://www.fwi.co.uk/business/pay-survey-who-gets-what-in-the-farm-industry
"Broken down further, the figures reveal that relatives of farm owners are not drawing full salaries. The average recorded wage for these people was just £12,616. The highest wages were for farm managers, who earned roughly £10,000 more than the average wage of £25,565.
The biggest earners in related businesses were technical managers at £40,111 – above the average of £34,158."
| | | |
| If MT brings zero productivity gain... | Aug 14, 2019 |
...then there is no point using it.
My advice: use MT by all means if it makes economic sense.
And I agree that many translators use CAT tools' MT plug-ins without admitting it.
Only you can tell whether MT increases your productivity. Sure, MT is gaining ground, but the amount of disposable, MT-prepped, run-of-the-mill, social or invisible content keeps increasing.
I have assessed and incorporated productivity tools in my work, from DNS (typing is indeed bo... See more ...then there is no point using it.
My advice: use MT by all means if it makes economic sense.
And I agree that many translators use CAT tools' MT plug-ins without admitting it.
Only you can tell whether MT increases your productivity. Sure, MT is gaining ground, but the amount of disposable, MT-prepped, run-of-the-mill, social or invisible content keeps increasing.
I have assessed and incorporated productivity tools in my work, from DNS (typing is indeed boring) and TO3000 to MemoQ and Copernic. If it made my life easier or boosted my productivity, I would have enabled MT plug-ins. But although there is a humongous amount of bilingual data around in my language pair, it doesn't help. And I hate this small lag that MT plug-ins require.
I have also post-edited hundreds of thousands of words of very usable raw MT with an agency using customised in-house MT systems a decade or so ago. Although properly paid, I can't sustain this activity for long: I can't get bored stiff every minute of my working day on an interesting translation.
The good news for dinosaurs is that agencies have end clients who still require human-only input. And not the least in technical subjects, where most seem to think that PEMT and a translation written by a decent translator are equal in readibility or usability, whatever the subject as soon as it is "technical".
To me, MTPE lacks the fundamental element for which I like my job: you don't "build" your own knowledge and understanding of whatever text you are working on. You don't "ghost-follow" a process or a reasoning. You're presented with a tentative translation that you must validate or correct. The mindset is totally different, and certainly not as engaging, gratifying or pleasant as learning new things.
MTPE is also about compromising, because the more edits, the less productivity (therefore money). Post-editors must find the right trade-off: make only the changes needed to keep their business sustainable, while approaching human-only quality to keep their customers. All that considering the ultimate goal: PEMT MUST require less time than human translation. Quite a balancing act!
Of course, economic constraints also apply to human translation, but at least the first translation attempt is mine, so I don't have to check meaning, consistency, connections, relevance, terms, flow or logic at the rereading stage, because I have read and understood everything that's in it like a human being.
Gary Evans wrote:
Of course all translators need to have good writing skills.
Wishful thinking. Economic constraints, poor business skills, hidden use of MT or incompetence spoil the picture.
And turning to MTPE will not help them improve!
Human translators promote their added value as human beings who understand what they read and write in each language. I seem to see that the term "translation", which used to refer to "quality human translation", is being replaced with "transcreation" to make a difference with PEMT. "translation" may soon be a synonym for "post-edited machine translation" and "transcreation" for "quality human translation".
Philippe ▲ Collapse
| | | |
| The bottom line of what I wrote earlier | Aug 14, 2019 |
Hello B.D Finch,
Thank you for your reply and sharing your thoughts about my earlier comment. With that said, I feel I need to re-clarify what I said earlier.
The example I gave about agricultural sector was just a likening, a metaphor if you'd like. Of course, we can hardly expect a process that occurred nearly a 1.5 century ago to reoccur in the exact same manner. But, I want to point out one part of your comment. You said, it drove out the "amateur farmers". Now I i... See more Hello B.D Finch,
Thank you for your reply and sharing your thoughts about my earlier comment. With that said, I feel I need to re-clarify what I said earlier.
The example I gave about agricultural sector was just a likening, a metaphor if you'd like. Of course, we can hardly expect a process that occurred nearly a 1.5 century ago to reoccur in the exact same manner. But, I want to point out one part of your comment. You said, it drove out the "amateur farmers". Now I invite you to think, don't we also have "amateur translators" in our industry? The very people, who claims that they can call themselves "translators" just because they think they know the language even though they are not educated properly. Those are the same people knocking down the prices in a non-ethical competitive way just in order to get the job, which then leads to the fact that employers start offering ridiculous and insulting translation prices to us translators. Personally I would not mourn if those kind of people leave our industry and start working at somewhere else, which will eventually give us a more professional sector. And if technology is the one to realize this much-needed natural selection, so be it.
I would refrain from discussing the ethics of farming or animal husbandry under this particular subject. Let it be a subject of another day.
I would like to thank you for the links you provided, but there is one last sentence that I'd like to draw attention on in your comment. You said the people that make the most money are "technical managers" . Aren't those people the ones, who are intelligent enough to cope with the technical and technological issues because they developed themselves in the knowledge of that particular areas? So I do think they deserve every penny of what they earn, because they know how to deal with technology and adapted themselves to the necessities of our era, unlike some others that take the easy way out and chose not to spend time on developing themselves in technological knowledge. ▲ Collapse
| | |
|
|
|
Daniel Williams
United Kingdom
Local time: 09:24
Member (2018)
German to English
+ ...
| If it's connected to the Internet, it's not confidential - just hidden. | Aug 14, 2019 |
I take the point that some MT providers state they do not store data, but as a veteran of many IT security briefings from my previous career, I still assume that any system that is connected to the Internet is not secure. A hacker only has to get lucky once. By the same token, anything you put on the Internet should be assumed to be publicly accessible for all eternity whether it has been 'deleted' or not; the only question is whether someone can be bothered to take the time to look for it.
... See more I take the point that some MT providers state they do not store data, but as a veteran of many IT security briefings from my previous career, I still assume that any system that is connected to the Internet is not secure. A hacker only has to get lucky once. By the same token, anything you put on the Internet should be assumed to be publicly accessible for all eternity whether it has been 'deleted' or not; the only question is whether someone can be bothered to take the time to look for it.
This is not to say that no-one should ever use MT; from the MTPE work I do I can see the machines are getting better all the time and they are a good option for certain text types. However, I still go by the general rule that if a system is connected to the Internet, it cannot be assumed that any data has that has ever gone through that system is (or will remain) confidential. Along with the fact that I am yet to see a machine that can actually write decent English that works for a target audience as opposed to translating exactly what's in front of it, this confidentiality aspect is one of the major reasons I cite when advising clients not to put their text through a machine.
[Edited at 2019-08-14 13:56 GMT] ▲ Collapse
| | | |
Jessica Noyes
United States
Local time: 04:24
Member
Spanish to English
+ ...
| Initial question is a springboard for comments in favor of MT | Aug 14, 2019 |
So my question is: who is not using MT and why?
Does the Asker really care what someone's reasons may be for not using MT? The pattern I am seeing is:
1. Translators respond with their opinions of, and experiences with, MT.
2. Asker addresses each answer individually, presenting arguments against one or more of their points, making remarks such as "Which MT programs have you tried?"
Personally, I find this approach annoying. A less misleading initial post would be, "Tell me why you are not using MT, and I will tell you why you are mistaken."
| | | |
Gary Evans
Germany
Local time: 10:24
German to English
TOPIC STARTER | But if MT helps... | Aug 14, 2019 |
...then there is a pointing using it.
Hi Philippe,
Yours in the first response which directly addresses my question. Thank you.
I disagree that one learns less when using CAT tools. You still have to do the research on the topic. But still, as with any sort of automation, the operator tends to become less capable over time. Take fly-by-wire technology. Sometimes the pilot must intervene and fly the plane out of trouble. If they're out of practice, they're l... See more ...then there is a pointing using it.
Hi Philippe,
Yours in the first response which directly addresses my question. Thank you.
I disagree that one learns less when using CAT tools. You still have to do the research on the topic. But still, as with any sort of automation, the operator tends to become less capable over time. Take fly-by-wire technology. Sometimes the pilot must intervene and fly the plane out of trouble. If they're out of practice, they're less capable of doing this. Just wait until self-driving cars are mainstream and see how people cope when the system demands they take back control.
"And I agree that many translators use CAT tools' MT plug-ins without admitting it."
And that's the point I'm trying to make. If it helps, then use it, but be careful when working with agencies, as many forbid the use of MT, mainly due to data security reasons.
Regards,
Gary
Philippe Etienne wrote: ...then there is no point using it. My advice: use MT by all means if it makes economic sense. And I agree that many translators use CAT tools' MT plug-ins without admitting it. Only you can tell whether MT increases your productivity. Sure, MT is gaining ground, but the amount of disposable, MT-prepped, run-of-the-mill, social or invisible content keeps increasing. I have assessed and incorporated productivity tools in my work, from DNS (typing is indeed boring) and TO3000 to MemoQ and Copernic. If it made my life easier or boosted my productivity, I would have enabled MT plug-ins. But although there is a humongous amount of bilingual data around in my language pair, it doesn't help. And I hate this small lag that MT plug-ins require. I have also post-edited hundreds of thousands of words of very usable raw MT with an agency using customised in-house MT systems a decade or so ago. Although properly paid, I can't sustain this activity for long: I can't get bored stiff every minute of my working day on an interesting translation. The good news for dinosaurs is that agencies have end clients who still require human-only input. And not the least in technical subjects, where most seem to think that PEMT and a translation written by a decent translator are equal in readibility or usability, whatever the subject as soon as it is "technical". To me, MTPE lacks the fundamental element for which I like my job: you don't "build" your own knowledge and understanding of whatever text you are working on. You don't "ghost-follow" a process or a reasoning. You're presented with a tentative translation that you must validate or correct. The mindset is totally different, and certainly not as engaging, gratifying or pleasant as learning new things. MTPE is also about compromising, because the more edits, the less productivity (therefore money). Post-editors must find the right trade-off: make only the changes needed to keep their business sustainable, while approaching human-only quality to keep their customers. All that considering the ultimate goal: PEMT MUST require less time than human translation. Quite a balancing act! Of course, economic constraints also apply to human translation, but at least the first translation attempt is mine, so I don't have to check meaning, consistency, connections, relevance, terms, flow or logic at the rereading stage, because I have read and understood everything that's in it like a human being. Gary Evans wrote:
Of course all translators need to have good writing skills. Wishful thinking. Economic constraints, poor business skills, hidden use of MT or incompetence spoil the picture. And turning to MTPE will not help them improve! Human translators promote their added value as human beings who understand what they read and write in each language. I seem to see that the term "translation", which used to refer to "quality human translation", is being replaced with "transcreation" to make a difference with PEMT. "translation" may soon be a synonym for "post-edited machine translation" and "transcreation" for "quality human translation". Philippe ▲ Collapse
| | | |
Dan Lucas
United Kingdom
Local time: 09:24
Member (2014)
Japanese to English
| Irrelevant until clients agree | Aug 14, 2019 |
Gary Evans wrote:
I appreciate concerns about confidentiality, but (naming no names) there are MT engines which explicitly do not store any data as part of their offer.
Your opinion (and indeed my own opinion) is neither here nor there. MT providers don't have to convince me, they have to convince my clients. I am aware that some providers claim to no longer peek at the data. The question for my clients is how far they can trust such promises. For material non-public information of the kind my clients deal with, I can understand why they prefer to err on the side of caution.
Regards,
Dan
| | |
|
|
|
Gary Evans
Germany
Local time: 10:24
German to English
TOPIC STARTER | The Asker cares | Aug 14, 2019 |
Hi Jessica,
The Asker has a name and my question was genuine, but you didn't answer it.
"A less misleading initial post would be, "Tell me why you are not using MT, and I will tell you why you are mistaken." "
I've read numerous good reasons for not using MT on this thread. It's not suitable for fashion, poetry, literature etc. I won't dream of promoting MT in these fields. So your proposed alternative question doesn't apply here. If the main reason is data... See more Hi Jessica,
The Asker has a name and my question was genuine, but you didn't answer it.
"A less misleading initial post would be, "Tell me why you are not using MT, and I will tell you why you are mistaken." "
I've read numerous good reasons for not using MT on this thread. It's not suitable for fashion, poetry, literature etc. I won't dream of promoting MT in these fields. So your proposed alternative question doesn't apply here. If the main reason is data security, then I repeat: some MT engines do not store the data. Of course you can become paranoid about people somehow stealing any data that is on the Internet, but that's really just paranoia.
I asked who is not using MT to find out how many are actually using it without the knowledge of the client/agency. I'd appreciate honest answers rather than nitpicking.
Regards,
Gary
Jessica Noyes wrote: So my question is: who is not using MT and why?
Does the Asker really care what someone's reasons may be for not using MT? The pattern I am seeing is: 1. Translators respond with their opinions of, and experiences with, MT. 2. Asker addresses each answer individually, presenting arguments against one or more of their points, making remarks such as "Which MT programs have you tried?" Personally, I find this approach annoying. A less misleading initial post would be, "Tell me why you are not using MT, and I will tell you why you are mistaken."
[Edited at 2019-08-14 15:00 GMT] ▲ Collapse
| | | |
Gary Evans
Germany
Local time: 10:24
German to English
TOPIC STARTER | Then it's about trust | Aug 14, 2019 |
Hi Dan,
If a machine translation provider offers confidentiality as part of their service, you'd think that this would be the case. That stated, recent issues with contractors listening in to Alexa and Siri output do damage to the business models of the corporations concerned. Of course this is all about trust and I can understand clients erring on the side of caution.
FYI I was recently offered a big translation, which would have been suitable for the MT engine I use. ... See more Hi Dan,
If a machine translation provider offers confidentiality as part of their service, you'd think that this would be the case. That stated, recent issues with contractors listening in to Alexa and Siri output do damage to the business models of the corporations concerned. Of course this is all about trust and I can understand clients erring on the side of caution.
FYI I was recently offered a big translation, which would have been suitable for the MT engine I use. I suggested using this and the agency replied with words to the effect of "use whatever tools you need, but the responsibility lies with you". This despite the NDA explicitly forbidding the use of MT.
Regards,
Gary
Dan Lucas wrote: Gary Evans wrote:
I appreciate concerns about confidentiality, but (naming no names) there are MT engines which explicitly do not store any data as part of their offer. Your opinion (and indeed my own opinion) is neither here nor there. MT providers don't have to convince me, they have to convince my clients. I am aware that some providers claim to no longer peek at the data. The question for my clients is how far they can trust such promises. For material non-public information of the kind my clients deal with, I can understand why they prefer to err on the side of caution. Regards, Dan ▲ Collapse
| | | |
| Using CAT tools daily since 2000 | Aug 14, 2019 |
Gary Evans wrote:
...I disagree that one learns less when using CAT tools...
MT, not CAT tools. I find CAT tools helpful, but not their MT plug-ins.
Philippe
| | | |
| Pages in topic: < [1 2 3 4 5 6 7] > |
Login to reply/comment 
Translation news
