The MT wishful thinking Decalogue 论题张贴者: Claudio Porcellana (X)
|
|---|
Claudio Porcellana (X) 
意大利
How much time is needed to have MT tools really suitable for mere translators as CATs are?
So far they seem more suitable for semantics engineer or professors, and even MMM is not enough "mere" to me.
I mean MT tools that:
1. have a demonstrable performance better than GT
2. integrate smoothly with MS Office and most CATs, i.e. our regular work-flow
3. have an average learning curve comparable to most CATs
4. ha... See more How much time is needed to have MT tools really suitable for mere translators as CATs are?
So far they seem more suitable for semantics engineer or professors, and even MMM is not enough "mere" to me.
I mean MT tools that:
1. have a demonstrable performance better than GT
2. integrate smoothly with MS Office and most CATs, i.e. our regular work-flow
3. have an average learning curve comparable to most CATs
4. have a reasonable cost for the base tool, and no usefulness features (as text-to-speech for example), i.e. one (1) language couple only and no specialistic dictionaries, but with unlimited entries for unlimited custom dictionaries
5. have unlimited entries for unlimited custom dictionaries
6. - have good specialist dictionaries in all languages couples (very few are EN to IT, for example)
7. don't need taking on Tianhe-1 to train them
8. as an alternative, the seller trains the MT with our own memories/corpora
9. can be tried as fully operational to judge if the game (quite time & money expensive) is worth the candle (only Prompt follows partly this policy)
10. as an alternative, the seller trains the MT with a reasonable part of our own memories/corpora, and/or makes available an on-line tool, but whose performance must be not less than GT
NOTE: I consider GT the gold standard as it's for free and it's "good enough" in skilled hands (personally tested in the field for over a month)
BTW, I'm a TAUS member and I downloaded so far the whole EN to IT TAUS corpora, plus I have many other authoritative references
I say that to explain that data are not a problem to me, a part of their quality that's a different kettle of fish, but the complexity of current MT is a major issue that forces me to stick on GT
and I quoted the quality as even DGT corpora contain blunders and I think that even the LSP's God isn't indeed able to proofread zillions of words
Claudio ▲ Collapse
| | | | | Depends on the task | Nov 26, 2010 |
MT tools may already be good enough for some very specific tasks, but I seriously doubt we'll see any general-purpose MT system within our lifetime that would be usable for serious high-quality work (let alone passing a Turing test). To put it into pespective, unmanned aircraft do have their own fields of application (like military reconnaissance), but unmanned passenger jets are not even on the agenda.
Every couple of years, I check out a few well-known MT tools. So far, all of th... See more MT tools may already be good enough for some very specific tasks, but I seriously doubt we'll see any general-purpose MT system within our lifetime that would be usable for serious high-quality work (let alone passing a Turing test). To put it into pespective, unmanned aircraft do have their own fields of application (like military reconnaissance), but unmanned passenger jets are not even on the agenda.
Every couple of years, I check out a few well-known MT tools. So far, all of them have been failing miserably. If you need a high-quality translation, the time required to proofread a machine-translated text remains comparable to the time required to translate it from scratch. Alternatively, for similar texts within a given field, one may possibly be able to produce decent results, but the cost of training the system easily offsets the benefits of MT.
By the way, even CATs have a long way to go in terms of usability. Unfortunately, the situation is not really getting any better. In my opinion, Trados used to have the best interface ergonomics in the industry, but given how kludgy their Studio 2009 is, I am not holding my breath for any significant improvements. Much-touted Wordfast is a total usability disaster. The list can go on. ▲ Collapse
| | | | Michael Davies
丹麦
Local time: 17:12
正式会员 (自2009)
Danish丹麦语译成English英语
+ ...
| GT: Google Translate or GoldText Translator or .... ? | Nov 26, 2010 |
What do YOU mean by GT?
It is usually a good idea to inform readers what is meant by an abbreviation ... just to avoid incorrect interpretation.
BTW (By the way) I use Wordfast (Classic, Pro and - increasingly -Anywhere), which I find to be a satisfactory (if not perfect) CAT (Computer Aided Translation) tool.
Wordfast Anywhere is, BTW, free of charge to use and has access to a VLTM (Very Large Translation Memory), which can make translation more efficient... See more What do YOU mean by GT?
It is usually a good idea to inform readers what is meant by an abbreviation ... just to avoid incorrect interpretation.
BTW (By the way) I use Wordfast (Classic, Pro and - increasingly -Anywhere), which I find to be a satisfactory (if not perfect) CAT (Computer Aided Translation) tool.
Wordfast Anywhere is, BTW, free of charge to use and has access to a VLTM (Very Large Translation Memory), which can make translation more efficient by reducing the time required to translate a text without just handing the whole job over to a machine.
Regards,
Michael Davies,
MJD Translating,
Denmark ▲ Collapse
| | | | Claudio Porcellana (X)
意大利
主题发起人 | The MT wishful thinking Decalogue | Nov 26, 2010 |
@Michael
GT = Google Translator
sorry I forgot mentioning it
anyway, considering that MT (machine translation) vendors want selling their tools, and many MT gurus think (correctly) that a huge gap still exist between MT vendors (and MLV [Multi Language Vendors] that use MTs) and most translators, I think that sellers must do a similar huge effort to entice us
@Anton
even a simple... See more @Michael
GT = Google Translator
sorry I forgot mentioning it
anyway, considering that MT (machine translation) vendors want selling their tools, and many MT gurus think (correctly) that a huge gap still exist between MT vendors (and MLV [Multi Language Vendors] that use MTs) and most translators, I think that sellers must do a similar huge effort to entice us
@Anton
even a simple tool as GT is very useful in skilled hands, but a possible boomerang in the opposite situation
this is why I claim that "MT + DVM + common sense sparkle!", NOT clearly "MT + monkey sparkle!"
(DVM = Abbreviation for Doctor of Veterinary Medicine)
the matter is that you must learn first how to edit a MT ouput, that's a relly different kettle of fish from editing a human translation ouput
as a peer says, http://bit.ly/cNakwA, a post editor should not re-translate, but simply move translated words in their correct position in the sentence, make sure plurals and genders of words are concordant within a sentence, check that word declension has been adhered to, and that terminology is correct and consistent
and if you stick to these rules the picture changes a lot
clearly, I don't translate everything in that way: I use this work-flow for low budget tasks, or when a customer is in a hurry and agree I use it
Claudio
P.S. I'm a new Wordfast fan: too sad that it cannot translate ttx files that aren't pre-segmented by TagEditor ...
[Modificato alle 2010-11-26 11:07 GMT] ▲ Collapse
| | |
|
|
|
Jeff Allen
法国
Local time: 17:12
Multiplelanguages多语种
+ ...
| Google Translate vs other MT solutions | Nov 26, 2010 |
On the topic of Google Translate (GT) and how it fits within the overall picture and landscape of MT tools, colleagues and I have addressed this specific topic in each of the MT panels at the recent series of ProZ Virtual conferences below. GT is not and should not be "the reference" to MT for the reasons stated in those sessions. It is just one of many, and just happens to be the most visible on... See more On the topic of Google Translate (GT) and how it fits within the overall picture and landscape of MT tools, colleagues and I have addressed this specific topic in each of the MT panels at the recent series of ProZ Virtual conferences below. GT is not and should not be "the reference" to MT for the reasons stated in those sessions. It is just one of many, and just happens to be the most visible one.
http://www.proz.com/translation3/
- go to the relevant tab (Freelancers or Translation Agencies or Globalization Translation consumers)
- click on the sessions panel and choose MT
Machine Translation (MT) trends Panel session : ProZ Virtual
Conference for Agencies for-freelancers (30 Sept 2010)
Machine Translation (MT) trends Panel session : ProZ Virtual
Conference for Translation Agencies : (13 Oct 2010)
Machine Translation (MT) trends Panel session : ProZ Virtual
Conference for Globalization Translation Consumers (10 Nov 2010)
As I've mentioned in various posts here on ProZ and in one or several of those sessions, a free coffee distributor is not the reference for best coffee, nor is a 1-speed bicycle the reference point for doing a Tour de France bike race, not is inexpensively distributed camera or phone to be the reference to professional photographers or business people. It needs to be placed in context. GT is intended to handle the general audience, general topic, content gisting public. It is not feature-designed to cover the needs of professional translators.
In addition to those panel sessions, I also specifically cover the topic of Google Translate in comparision with other MT solution offers in the set of MT webinars offered here on ProZ. The links to those are in my bio available through the various panel sessions mentioned above.
- Overview on machine translation
- MT for freelancers webinar
- MT for Translation agencies webinar
also a presentation on this (in English and/or French, depending on the audience) at the annual ProZ conference in Paris in 4 Dec 2010
http://www.proz.com/virtual-conferences/139/program/4979
+ an open Q&A session on it (in French) on 3 Dec 2010
Explaining how GT compares with other MT solutions, and how the lack of configurability and customization of GT influences and heavily increases the downstream workload of post-editing are among the topics I try to address in detail. These are key items in helping this audience make decisions to adopt MT or not.
Jeff ▲ Collapse
| | | | Claudio Porcellana (X)
意大利
主题发起人 | The MT wishful thinking Decalogue | Nov 26, 2010 |
Hi Jeff
I read your documents, but you are still speaking of "ethereal stuff", while I'm really a doubting Thomas ...
I think that if MT vendors will not deliver fully functional demos, and will not try to sell really useable tools, at least as useable as most CATs, the dialogue will remain a dialogue of the deaf and that gap will remain for eons
and GT is a gold standard to me as it is for f... See more Hi Jeff
I read your documents, but you are still speaking of "ethereal stuff", while I'm really a doubting Thomas ...
I think that if MT vendors will not deliver fully functional demos, and will not try to sell really useable tools, at least as useable as most CATs, the dialogue will remain a dialogue of the deaf and that gap will remain for eons
and GT is a gold standard to me as it is for free!
how can a MT vendor ask me to buy its tool if there is no demo and/or the demo is unfriendly and/or it's performances are worst than a free tool???
this is greatly summarized by the Italian motto "come un serpente che si morde la coda"
(the English translation "it's a vicious circle" doesn't translate the same beauty, sorry )
Claudio ▲ Collapse
| | | | Claudio Porcellana (X)
意大利
主题发起人 | The MT wishful thinking Decalogue | Nov 26, 2010 |
BTW, following your reasoning Jeff
what's the point asking to gurus/visionaries etc. if their dreams/visions are realizable?
I want solid stuff sorry
and it's not smart to me that MT vendors aren't able to understand this simple concept, because narrowing the gap between "MT and translators that believe in MT", rather than between "MT and peers that hate MT", should be their easier and ... See more BTW, following your reasoning Jeff
what's the point asking to gurus/visionaries etc. if their dreams/visions are realizable?
I want solid stuff sorry
and it's not smart to me that MT vendors aren't able to understand this simple concept, because narrowing the gap between "MT and translators that believe in MT", rather than between "MT and peers that hate MT", should be their easier and more profitable task, don't you think so?
Claudio ▲ Collapse
| | | | | We probably work in totally different domains | Nov 27, 2010 |
Claudio, if the procedure of "not re-translate, but simply move translated words in their correct position in the sentence, make sure plurals and genders of words are concordant within a sentence, check that word declension has been adhered to, and that terminology is correct and consistent" were applicable to everything I managed to get out of MT, I would be of a much higher opinion of it. Yes, Google may be good enough for easy texts (e.g. correspondence), especially between closely related la... See more Claudio, if the procedure of "not re-translate, but simply move translated words in their correct position in the sentence, make sure plurals and genders of words are concordant within a sentence, check that word declension has been adhered to, and that terminology is correct and consistent" were applicable to everything I managed to get out of MT, I would be of a much higher opinion of it. Yes, Google may be good enough for easy texts (e.g. correspondence), especially between closely related languages. I do occasionally use it for my own personal needs. However, from my own experience with the highly specialized texts I usually translate, more often than not Google produces a total gibberish, where half of the text is to be thrown away and only half can be edited in the way you describe.
[Edited at 2010-11-27 13:49 GMT] ▲ Collapse
| | |
|
|
|
I used ALS’s SmartMATE and had good experience with it. I was surprised by the high quality of the output – once I had cleaned my TMs. Being a newcomer to the world of MT, I liked the fact that there was no need to customise the service or indicate any settings whatsoever. The output needed post-editing, of course, but not much.
| | | | | Free vs non-free | Dec 20, 2011 |
Jeff Allen wrote:
As I've mentioned in various posts here on ProZ and in one or several of those sessions, a free coffee distributor is not the reference for best coffee, nor is a 1-speed bicycle the reference point for doing a Tour de France bike race, not is inexpensively distributed camera or phone to be the reference to professional photographers or business people. It needs to be placed in context. GT is intended to handle the general audience, general topic, content gisting public. It is not feature-designed to cover the needs of professional translators.
But this logic is flawed. What is the gold standard for search? We would easily say Google, which is free. What is the gold standard for music typesetting? I would confidently say Lilypond, which is also free. The gold standard for mathematical typesetting is TeX. Being free does not prevent something from having the quality of a gold standard.
By the way, for certain language pairs (such as EN>ZH), GT is not usable even for a general audience, and I remain unconvinced that the situation will be significantly different with specialized MT tools.
| | | | | High quality MT is possible | Jan 4, 2012 |
Ambrose Li wrote:
By the way, for certain language pairs (such as EN>ZH), GT is not usable even for a general audience, and I remain unconvinced that the situation will be significantly different with specialized MT tools.
Hi Ambrose,
I agree with Jeff - GT should not be a measuring stick for MT. There is technology today available that can give near-human quality MT results when fully customized.
As an example, this link (http://www.youtube.com/watch?feature=player_detailpage&v=hjK17GWynoU#t=1525s) provides a real world example for EN-ZH where our client Sajan customized an Asia Online Language Studio MT engine for their client (a major multinational IT company).
LSP:
Sajan
End Client Profile:
Large global multinational corporation in the IT domain.
Has developed its own proprietary MT system that has been developed over many years.
Language Pair:
English -> Simplified Chinese.
Domain:
IT
Quality:
Client performed their own metrics (both human and automated)
Asia Online Language Studio™ was 5 BLEU points better than the clients own MT solution.
Significant quality improvement after providing feedback – 65 BLEU score.
MT scored better than first pass human translation as per client’s feedback
Result:
Client extremely impressed with result especially when compared to the output of their own MT engine.
Client has commissioned Sajan to work with more languages
60% Cost Saving
70% Time Saving
[Zmieniono 2012-01-04 19:10 GMT]
| | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » The MT wishful thinking Decalogue | TM-Town | Manage your TMs and Terms ... and boost your translation business
Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.
More info » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
