| Pages in topic: < [1 2 3 4 5 6 7] > | '[MT] is most often used alongside [TM] as an adjunct to human translation'. Are you using it? Thread poster: Henry Dotterer
|
|---|
Eric Hahn (X) 
France
Local time: 08:43
French to German
+ ...
Giovanni Guarnieri MITI, MIL wrote:
recently, I tried SDL's automated translation feature available in Workbench (8.3.0.863 - Build 863) for a repetitive technical manual. I was quite surprised.
Is it a server-based feature, like in Babylon ?
| | | |
Eric Hahn wrote: Giovanni Guarnieri MITI, MIL wrote:
recently, I tried SDL's automated translation feature available in Workbench (8.3.0.863 - Build 863) for a repetitive technical manual. I was quite surprised.
Is it a server-based feature, like in Babylon ?
it's indeed server-based, but I know nothing about Babylon, so I can't compare the two. It's a beta release, at the moment...
| | | | Eric Hahn (X)
France
Local time: 08:43
French to German
+ ...
| Server-based | Mar 17, 2009 |
or like Google-Translate if you prefer 
| | | | Paul Cohen
Greenland
Local time: 05:43
German to English
+ ...
| Translating the message behind the words | Mar 17, 2009 |
I mostly do marketing and journalistic translations from German into English. It's highly creative work. I often don't just translate the words -- I translate the message behind the words. I cannot imagine a machine being able to do that.
The day they invent a machine that can, I'll eat my mouse pad.
| | |
|
|
|
Marina Soldati
Argentina
Local time: 03:43
Member (2005)
English to Spanish
+ ...
| Different experience | Mar 17, 2009 |
Giovanni Guarnieri MITI, MIL wrote:
recently, I tried SDL's automated translation feature available in Workbench (8.3.0.863 - Build 863) for a repetitive technical manual. I was quite surprised. Sometimes (admittedly not many) the translation would be perfect, other times I had to edit it (a bit or heavily), many times it was useless, but I believe I saved some time by using it. The terminology was correct in some instances... It can be useful along your TM...
G
I tried SDL automated translation (EN>ES) with two types of texts:
A techncial text, where 95% of the automatically translated segments had to be heavily edited, there were some mispellings too. A real waste of time.
A general text, taken from an online newspaper. Google Translate produced a much better translation than SDL.
For me, at least at this stage, SDL automated translation is a waste of time (and money).
I haven´t tried any other MT software.
Regards,
Marina
Edited for typo!
[Edited at 2009-03-17 13:42 GMT]
| | | |
Marina Pastora Soldati wrote: Giovanni Guarnieri MITI, MIL wrote:
recently, I tried SDL's automated translation feature available in Workbench (8.3.0.863 - Build 863) for a repetitive technical manual. I was quite surprised. Sometimes (admittedly not many) the translation would be perfect, other times I had to edit it (a bit or heavily), many times it was useless, but I believe I saved some time by using it. The terminology was correct in some instances... It can be useful along your TM...
G I tried SDL automated translation (EN>ES) with two types of texts: A techncial text, where 95% of the automatically translated segments had to be heavily edited, there were some mispellings too. A real waste of time. A general text, taken from an online newspaper. Google Translate produced a much better translation than SDL. For me, at least at this stage, SDL automated translation is a waste of time (and money). I haven´t tried any other MT software. Regards, Marina Edited for typo! [Edited at 2009-03-17 13:42 GMT]
no typos in Italian... maybe some MTs are more careful than others...
[Edited at 2009-03-17 14:30 GMT]
| | | | Vito Smolej
Germany
Local time: 08:43
Member (2004)
English to Slovenian
+ ...
SITE LOCALIZER | I used proMT | Mar 17, 2009 |
... for some low-brow work (Email quatsch, minutes etc) in DEEN pair. The results were way beyond the babelfish and google level but of course required some nudging.
More I cant say - these languages as targets are off-limit for me anyhow and the hell will freeze over before Slovenian gets included (same applies to DNS too for instance).
Regards
Vito
[added]
An example (proMT Lite - I (mis)use Kleist, so that the original does not need ... See more ... for some low-brow work (Email quatsch, minutes etc) in DEEN pair. The results were way beyond the babelfish and google level but of course required some nudging.
More I cant say - these languages as targets are off-limit for me anyhow and the hell will freeze over before Slovenian gets included (same applies to DNS too for instance).
Regards
Vito
[added]
An example (proMT Lite - I (mis)use Kleist, so that the original does not need to be quoted):
"On the shores Havel lived, at the middle of the sixteenth century, a horse dealer, called Michael Kohlhaas, son of a school master, one of the most upright and most gruesome people at the same time of his time. - This extraordinary man would become, till his thirtieth year could be valid for the pattern of a good citizen. He owned in a village which leads the name still from him, a Meier's court on which he lived by his trade quietly; he educated the children whom his woman gave him, in the fear of God, to the industriousness and loyalty; not one was among his neighbours who would not have taken pleasure of his charity, or his justice; briefly, the world would have had to bless his memory if he had not gone to extremes in a virtue. However, the legal feeling made him the robber and murderer."
[Edited at 2009-03-17 19:49 GMT] ▲ Collapse
| | | | | I use it reasonably | Mar 17, 2009 |
I don't see anything wrong with reasonable use of GT for example.
By contrast, I would consider indiscriminate batch processing of whole documents through GT unreasonable.
I use software which lets me send a single segment/sentence to GT, so I decide whether to MT or not to MT, and on top of that I wrote my own plugin for it which lets me edit out any sensitive inf... See more I don't see anything wrong with reasonable use of GT for example.
By contrast, I would consider indiscriminate batch processing of whole documents through GT unreasonable.
I use software which lets me send a single segment/sentence to GT, so I decide whether to MT or not to MT, and on top of that I wrote my own plugin for it which lets me edit out any sensitive information before the unit is sent to GT.
GT is surprisingly good at what I'd call "eurolegalese" and similar texts and quite good in some other areas. Some editing is necessary, but it is by no means heavy editing.
So, used with care and reason MT can help and there is no reason for refraining from using it while it's free, and fits the type of texts you translate.
MT based on statistical methods is here and ignoring it won't make it go away.
Regards,
Piotr ▲ Collapse
| | |
|
|
|
| MT on the horizon | Mar 17, 2009 |
Piotr Bienkowski wrote:
I don't see anything wrong with reasonable use of GT for example.
By contrast, I would consider indiscriminate batch processing of whole documents through GT unreasonable.
MT based on statistical methods is here and ignoring it won't make it go away.[/quote]
This is the impression I got, too. I've been reading about MT a lot lately, incidentally because of a very interesting thread about this there was here in Proz.com a few months ago, but also because of some lectures I attended about this during a recent conference on technical documentation.
So, in order to learn more about this, I got in contact with an industry expert who kindly accepted to share his views about this in my blog (everyone is invited to read it by clicking here).
Of course MT is not nearly perfect right now, but it is being slowly but surely refined/improved, so it is definitely on the horizon for some very specific cases of translation.
Saludos,
Ivette
P.S.: Oops, I forgot, to answer the question of this thread, I am not using it right now. I used it in some specific case about a year or more ago, but the result was not good for the kind of manual it was used. I do use, for example, Google's translation tools to look up words.
[Edited at 2009-03-17 23:35 GMT]
| | | | Fabio Descalzi
Uruguay
Local time: 03:43
Member (2004)
German to Spanish
+ ...
| Fan of TM, but don't use MT at all | Mar 17, 2009 |
TM is a great companion that increases productivity. But... when it's my baby
I strongly distrust TMs given to me by third parties; whenever I have to use others' TMs, I carefully review sentence by sentence.
Regarding MT, all I can say is that I tried to proofread some MT-translated texts... and it made me sick to find too much nonsense...
| | | | | Most of the time, it's garbage. But when it works, it's awsome. | Mar 18, 2009 |
Most of the texts I translate don't result in useful sentences. For example, technical material.
However, I find it works very well with small TMs, as most terminology is not yet established and I can make good use of the terms suggested by the MT engine. In particular, in texts of a general nature, like medical benefits and healthcare, I have doubled my productivity and usually surpass 1000 source words per hour, even when applying lots of changes to the suggested terminology and t... See more Most of the texts I translate don't result in useful sentences. For example, technical material.
However, I find it works very well with small TMs, as most terminology is not yet established and I can make good use of the terms suggested by the MT engine. In particular, in texts of a general nature, like medical benefits and healthcare, I have doubled my productivity and usually surpass 1000 source words per hour, even when applying lots of changes to the suggested terminology and turning phrases around.
If you'd like more details, see my posting in "How big of a threat is Google translate?" ( http://www.proz.com/post/1049413#1049413 ).
Daniel
[Edited at 2009-03-18 00:34 GMT] ▲ Collapse
| | | | Jeff Allen
France
Local time: 08:43
Multiplelanguages
+ ...
| Henry, here is a link on a ProZ forum about MT preparation steps | Aug 29, 2009 |
NR_Stedman wrote:
Jeff Allen who is a recognized expert on MT helped me make my choice in 2005. There were then two competing systems, Reverso and Systran...
it is good to hear from you again. Glad to know that the few exchanges we had on dictionary import (from what I recall off-hand) were able to get you using MT in a productive way.
Henry D wrote:
Thanks for the info! I know Jeff Allen writes quite a bit online. Is there a link that was particularly helpful to you on this topic?
Hi Henry,
Try this post on ProZ that I looked up again this evening in replying to another request on the same topic in another forum here.
MT preparation steps
http://www.proz.com/forum/proofreading_editing_reviewing/133180-machine_translation_postediting_translators_views_needed.html#1123567
Jeff
[Edited at 2009-08-30 00:17 GMT]
[Edited at 2009-08-30 00:18 GMT]
| | |
|
|
|
Jeff Allen
France
Local time: 08:43
Multiplelanguages
+ ...
| I use MT all the time | Aug 30, 2009 |
Lots of posts here on ProZ on that topic.
If I can just reply to comments made by others above in this thread.
@ Vito, you mention ProMTLite, which means you are not using the full-blown version (Promt Expert) with customization functions, which NR Stedman mentions in his post, and which I mention quite often. If you aren't doing any dictionary customization, then you'll get generic MT output, similar to as if you ran the translation through the MT engine online at the... See more Lots of posts here on ProZ on that topic.
If I can just reply to comments made by others above in this thread.
@ Vito, you mention ProMTLite, which means you are not using the full-blown version (Promt Expert) with customization functions, which NR Stedman mentions in his post, and which I mention quite often. If you aren't doing any dictionary customization, then you'll get generic MT output, similar to as if you ran the translation through the MT engine online at the ProMT website.
@ Heike, Astrid, German as target language has had its challenges, and it has often shown to be most successful on well customized projects, not the out of the box translation system. I know about a number of projects with German where it has been quite successful, and because the users have spent the appropriate amount of customization effort (notably dictionary development) to achieve the current quality. There are Germany companies that have successfully managed to implement MT over the long-term, and now Lucy Software not only offers MT solutions, but also with a speciality in SAP localization with several of their people having come from the dept at SAP that has implemented MT for 2 decades.
Heike, the lack of any customization work to train the system will end up causing much more postediting work. I've seen that with German language projects. Upfront dictionary investment effort leads to significant reduction in the postediting work, and I know this is the case for German.
MT postediting done on output from a non-customized MT system (something which I always advise to avoid doing) can be inefficient. I've even seen a project where the time effort was nearly double than regular translation costs. It could have been avoided, but those implementing the MT system didn't want to get trained on how to do it efficiently, so it cost them more in the end.
@ Marina, to my knowledge, the SDL automated translation system is the Transparent language MT system that SDL acquired in 2000-2001 timeframe, which they renamed as Autotrans for the desktop version and kept the Transparent MT name for the server version. Yet, SDL has kept its hands full on other technology developments (SDLX, acquisition of Trados, acquisition of Passolo, etc), so the investment in the MT system probably hasn't been as significant as they have definitely kept focused on these other technologies.
@ Tomás, the reference to the O'Brien article would most likely cite controlled language as a constraint, because Sharon O'Brien did her PhD on the combination of Controlled Language with MT.
@ Margreet, as for Dutch, it is one of the languages that does not seem to have had as much investment in the dictionary work until more recently. I believe that the Systran v6 MT system put a lot of improvement work into their Dutch language grammar and dictionary, but likely not as much as is regularly invested on French and Spanish, and likely Chinese (things like the Olympics can change where to invest effort to hit the widest coverage, that's just an educated guess)
@ Eric, actually there are some Japanese MT systems that have been well developed over a couple of decades, including necessary semantic level framework for them to handle JP-EN
And as for the MT implementation by Henkel ask for and get training on how to customize the system with their language specialists? If not, then woe to them.
I am aware of very successful implementations of abstract data sheets including in the chemical sector.
I might have missed replying to others, but wanted to at least cover the remarks that stuck out.
Hope that helps.
Jeff ▲ Collapse
| | | | Susan Welsh
United States
Local time: 02:43
Russian to English
+ ...
| How MT works | Aug 30, 2009 |
Piotr Bienkowski wrote:
GT is surprisingly good at what I'd call "eurolegalese" and similar texts and quite good in some other areas. Some editing is necessary, but it is by no means heavy editing.
This is not surprising, because of the way MT works, which is not by linguistic analysis--and therefore has nothing to do with the structure of a language--but by STATISTICAL analysis of how frequently such-and-such a word is followed by such-and-such another word. This is done by compiling vast databases of dual-language material that is readily available in digitized form, such as EU documents and documents from governments. This is the same reason that it would tend NOT to be good at documents (e.g., creative literature) that are not of this nature.
See previous thread:
http://www.proz.com/forum/translation_theory_and_practice/135200-fyi:_article_on_machine_translation.html
There have been other recent threads on this topic that are relevant to the issues raised here.
I have used it once or twice, getting mostly garbage. But I am considering trying out Promt, after Viktoria Gimbe's posts on that. That is a system that, as others have noted here, you TRAIN to suit your own needs. It's not cheap, so I'm waiting until other tech priorities are past.
Susan
| | | | Eric Hahn (X)
France
Local time: 08:43
French to German
+ ...
Jeff Allen wrote:
@ Eric, actually there are some Japanese MT systems that have been well developed over a couple of decades, including necessary semantic level framework for them to handle JP-EN
And as for the MT implementation by Henkel ask for and get training on how to customize the system with their language specialists? If not, then woe to them.
I am aware of very successful implementations of abstract data sheets including in the chemical sector.
Jeff
I happen to know a former Henkel language specialist: He told me that they don't use Systran any more, that his department has been dissolved and that they outsource all japanese translation work.
[Modifié le 2009-08-30 13:40 GMT]
| | | | | Pages in topic: < [1 2 3 4 5 6 7] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » '[MT] is most often used alongside [TM] as an adjunct to human translation'. Are you using it? | 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 » |
| | Wordfast Pro | Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
