ProZ.com Freelancer Success Summit

Join ProZ.com/TV for a FREE online event April 7-9th! Empowering Freelance Translators & Interpreters to achieve their business objectives and reach their full potential 30+ hours of content, Live Q&A, Virtual Powwows, Access to Free Training, giveaways & more. Join 1,000's of linguists from around the globe at the ProZ.com/TV Freelancer Success Summit

Click for Full Participation
PROZ.COM COVID-19 RESOURCE CENTER
Access Covid-19 jobs, answer relevant terminology questions, read industry news and more.

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

Eleftherios Kritikakis  Identity Verified
United States
Local time: 12:53
Greek to English
+ ...
Will put agencies out of business Aug 30, 2009

Translation machines are cheap, cheaper than Trados.

So far, for Greek, they produce less than 10% correct results (only short sentences) which are out of context in most cases... more trouble than good.

When they get good enough, I will buy the Enterprise Version and use it.

As you can imagine, this could put translation agencies out of business. End clients will buy their own translation machine versions and then hire individual tech savvy editors from pr
... See more
Translation machines are cheap, cheaper than Trados.

So far, for Greek, they produce less than 10% correct results (only short sentences) which are out of context in most cases... more trouble than good.

When they get good enough, I will buy the Enterprise Version and use it.

As you can imagine, this could put translation agencies out of business. End clients will buy their own translation machine versions and then hire individual tech savvy editors from proz or any other source for editing/proofreading etc. at 1/4 of the price they would pay to the agency. An accurate machine eliminates the need for the specialized and complicated agency services, and requires only a tech savvy and reliable final editor (who also has other "friends" in the case of large documents).



[Edited at 2009-08-30 13:59 GMT]
Collapse


 

Jeff Allen  Identity Verified
France
Local time: 19:53
Multiplelanguages
+ ...
MT users in chemical sector & MT stat vs rule systems Aug 30, 2009

Comments to Eric and to Susan below

@ Eric:

Jeff Allen wrote:
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


Eric Hahn wrote:
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.


I left out a word "did" as in "did Henkel ask"

See Chemical Abstracts mentioned in two presentations by Systran

http://lirics.loria.fr/doc_pub/Barcelon21June05Systran.pdf
slide 25

http://www.lang-tech.org/2003/SPEAKERS_AND_PRESENTATIONS/PRESENTATIONS/SABATAKAKIS.PDF
slide 12

and Case studies from chemical and pharmaceutical domains presented by Systran at the International Chemical Information Conference. ICIC
http://www.infonortics.com/chemical/ch06/06chempro.html

Chemical Abstracts participated in a survey on MT users:
http://ice.he.net/~hedden/intro_mt.html

This link is an example of how not to use MT on abstract documents. Same mistake of using general online MT portals to try and translate technical terminology, instead of using a paid desktop or enterprise server version with dictionary customization.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2659868

There is mention of Henkel as an MT user many years ago
http://www.mt-archive.info/MTNI-1993-Vasconcellos.pdf


@Susan, unfortunately your statement below is not quite true.

Susan Welsh wrote:
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


Statistics-based MT is only one kind of MT system, and a fairly recent system provided by MT vendors.

a one page PPT overview of the different kinds of MT systems is in my presentation on Inbound versus Outbound Translation at Localization World 2004:
http://www.geocities.com/mtpostediting/LWBonn2004-A05JeffAllen.pdf
off the top of my head it is around slide 12.

Rule-based MT systems (based on linguistic grammar rules combined with dictionary entries with linguistic part of speech categorization) have been around for a few decades, and represent the majority of implementations at present.

The MT system used by the European Commission is a customized version of the Systran system sold to them in 1976. It is a linguistic-based (rule-based system), not at all a statistical-based system. they might be investigating statistical MT today, but that is not what they have implemented since the late 70s.

and in the following post on Proz, I went into more depth on the mixing of TM and MT technologies (with examples of which products) and how and when Statistical MT came in:
http://lit.proz.com/forum/translator_resources/100328-machine_translation:_your_experience_with_the_various_mt_programmes_state_of_play-page3.html#1000970

a more indepth explanation about statistical MT (like Google Translate) without getting too technical is at:
http://www.proz.com/forum/translator_resources/100328-machine_translation:_your_experience_with_the_various_mt_programmes_state_of_play-page2.html#998639

the ProMT software is quite good.

Hope that helps.

Jeff

[Edited at 2009-08-30 18:18 GMT]

[Edited at 2009-08-30 18:19 GMT]


 

Pablo Bouvier  Identity Verified
Local time: 19:53
German to Spanish
+ ...
Translation as a job will be dead in a short/mid term Aug 30, 2009

I think it does not matter, if we use MTs or any other CAT tools. With the globalization, paradigms have changed. But, we are not flexible and clever enough to adapt ourselves, considering the high speed to which they happen. Translation as a job will be dead in a short/mid term. But, we remain grasped to the "workplace concept ". A workplace that already does not exist anymore than in our imagination. Take only a look at the numbe... See more
I think it does not matter, if we use MTs or any other CAT tools. With the globalization, paradigms have changed. But, we are not flexible and clever enough to adapt ourselves, considering the high speed to which they happen. Translation as a job will be dead in a short/mid term. But, we remain grasped to the "workplace concept ". A workplace that already does not exist anymore than in our imagination. Take only a look at the number of unemployed people in European countries!

By every job that will be destroyed in Occident, they will be created some others in the East at a minor cost for the same productivity. And there is no doubt: European companies willing to survive will go offshore and make use of the East outsourcing competitive advantages. So, for all those who lives from a translation job like me, my recommendation is that they look for another job of extra high added value and that could not be easily exported to the East.

[Editado a las 2009-08-30 21:37 GMT]
Collapse


 

Speranza  Identity Verified

Local time: 19:53
Spanish to Russian
+ ...
Really? Aug 31, 2009

Pablo Bouvier wrote:

my recommendation is that they look for another job that could not be easily exported to the East.


Since when is translation into Spanish (or any other non-Eastern language for that matter) easy to export to the East?

Besides, it is irrelevant. I'll refrain from a lengthy, if useful, economics lecture, but you may benefit from reading up on the concept of comparative advantage.

A few figures to calm down the they-steal-our-jobs crowd: as of 2000, China, with a quarter of the world's population, produced less than 4% of the world's exports; Mexico, with a hundred million people and in free trade agreement with the U.S., exported less than Belgium; India, with a billion people, accounted for less than 1% of world exports. The poorest countries' situation is even worse: in 2000, they were responsible for 0.6% of everything all the major world traders imported, down from 0.9% in 1980. (Source: here.)

I suggest we take a break from the holy war against globalization and dedicate some effort to fighting ignorance at home instead.


 

Sushan Harshe
India
Local time: 23:23
English to Hindi
+ ...
MT's are still in beginning stage Aug 31, 2009

This is just an aggressive marketing strategy, below is an example of MT En>Hi

"Come home to take your food = आओ घर अपना खाना लेने के लिए"

This is very poooore literal translation. It has to walk, toooo long way to achieve at least understandable quality of language.

Elías Sauza... See more
This is just an aggressive marketing strategy, below is an example of MT En>Hi

"Come home to take your food = आओ घर अपना खाना लेने के लिए"

This is very poooore literal translation. It has to walk, toooo long way to achieve at least understandable quality of language.

Elías Sauza stated;

MT is far from being what some claim it is. I'd rather not use nor accept work that involves MT.


&

I agreeeeee!
Collapse


 

Pablo Bouvier  Identity Verified
Local time: 19:53
German to Spanish
+ ...
Really: Translation as a job will be dead in a short/mid term Aug 31, 2009

Nadejda Vega Cespedes wrote:

Since when is translation into Spanish (or any other non-Eastern language for that matter) easy to export to the East?


Since eastern countries will be more productive than western countries?

a) To use obsolete statistics of more or less 10 years ago, when the economic crisis began in August, 2007 (at least, according to semiofficial sources) does not seems to be quite founded.

b) Translation has nothing to do with export and import. It has to do with the supply and demand law. And, while demand falls down in stung in almost all western countries, it grows at equal speed or higher in almost all countries of emergent economies. Translation does not constitute any exception to the law of supply and demand, although it is an intangible product.

This is not of now. Simply, this trend has increased now. If you want an actualargument, not a theoretical one, or a book one: Ask yourself where catalonian textile or motorbikes industry has gone (Sangla, Derbi, Bultaco, Montesa, etc) and how it has affected to technical translation.

c) I am not taking any holy war against globalization, nor against job theft. It is, simply, the actual trend of the current economy. Nothing more.

And yes, you're right: We should take a break and dedicate some effort to fighting ignorance at home...

[Editado a las 2009-09-01 00:40 GMT]


 

Jeff Allen  Identity Verified
France
Local time: 19:53
Multiplelanguages
+ ...
Please be more explicit about MT brand, version, build, etc Aug 31, 2009

Sushan Harshe wrote:

Post title: MT's are still in beginning stage

This is just an aggressive marketing strategy, below is an example of MT En>Hi

"Come home to take your food = आओ घर अपना खाना लेने के लिए"

This is very poooore literal translation. It has to walk, toooo long way to achieve at least understandable quality of language.




Sushan,

Please be more explicit and show more rigour in info provided before making general blanket statements about the status of MT as a whole

EN > HI

The language direction is specified. That is a good start.
Now:
Which system? (Commercial, industry research system, academic research system?) Which version number of the system? (for an online system you usually won't get the version number, but the date tested would be minimal info).

Let's take a few examples:

Giovanni Guarnieri MITI, MIL wrote:
recently, I tried SDL's automated translation feature available in Workbench (8.3.0.863 - Build 863)


I refer to ProMT Pro and Expert versions 6/6.5/7 with the version depending on the language pair I am working with.
It is possible to look up the exact build numbers, but that is less critical for products that have been released on the market for a few years (ProMT version current at v8), usually more important when testing more recently released versions.

Why is this info important?

1) Different product models offer different features. The more you pay, the more features you get, just like any other product.
So if you say Systran Mobile Translator, Systran Home Translator, Systran Business Translator, Systran Office Translator, Systran Professional Translator, Systran Enterprise Server, SYSTRANLinks, SYSTRANet, there are sets of progressive features per product.
And yes it is possible that this can affect translation quality because if the high-end versions include specialized domain dictionaries that are not included in the lower-end products, then the translation simply won't be as good.

2) Commercially released, opensource, academic research....

If the system is an academic research project, then I simply don't have the same expectations as if it is a paid product with customer support.
is it a free online product to gather interest or a paid product intended to be a support tool for business tasks? This are simply different expectations.

* Academic research projects might be using free student labor to input dictionary entries and grammar rules
* Online free statistical MT systems might be limited to the corpus training data used to train the system. So if they focused on training the system on government policy documents, and you are trying to translate marketing content for the pharmaceutical domain, well that's a bit of a mismatch
* opensource systems are restricted to the volunteer help they get.
* commercially released products usually hire more skilled technicians to working on the systems, so there is a higher expectation of quality

3) maybe the version has reached end of life, and is no longer supported.

This is exactly what Adobe tells you for complaints raised about Adobe Reader. Go download the most recent released version, use it, and come back again if you have further problems.

The first thing that a technical support technician will ask for with regard to a complaint about any web browser is the version and build number.

Also take the analogy of Microsoft Office. If you were to complain about the compatability of files with your customers and about the features of the product, and then one asks you which version you are using, and you say "Microsoft Word 1997 SR-1", then the reply would probably be laughter with a response like "well, why don't you stop complaining and upgrade to Office 2007, or at least to Office 2003."

This is even similar to incidents raised about Translation Memory CAT tools. If you raise a support issue in a forum and then say you are using WordFast v3 or Trados v5, or Deja Vu v3, etc, then others will come back telling you to upgrade to fix a number of problems (I won't get into the debate over WFC and WFPro here, there are other threads for that topic).

4) not all language pairs have the same translation quality

It's normal because some languages have 3 decades of development behind them and others have 2 years, and others might have 6 months. Not every language direction (and I specific direction, not just pair) is at the same level of output quality. The more development invested over time, the higher likelihood that the quality will be improved.
So if this is a brand new offered language of Google Translate (a non commercial online product), then it is to no surprise that the quality is low. They are trying to prove a proof of concept to the public and to get feedback.

Once you provide brand, version, build info, then we have the minimal amount of info to determine why there might be issues with the output quality.

But your blanket statement above is like saying that you would never use a Canon camera again because Canon model A80 (year 2004) now in 2009 has a defect with the aperture setting, or you would never buy a Toyota car because the Toyota Corolla year 1999 (model XXXXXX) had a recall on the model in 2005 to fix a minor issue with the ABS brakes.

Thanks in advance for providing the additional info, and we'll see how that can orient a conclusion for your example.


Jeff

[Edited at 2009-08-31 21:26 GMT]

[Edited at 2009-08-31 21:28 GMT]


 

Samuel Murray  Identity Verified
Netherlands
Local time: 19:53
Member (2006)
English to Afrikaans
+ ...
SITE LOCALIZER
Strategies with Google Translate Sep 1, 2009

@Jeff:

Most of your articles that I've seen relate to using commercial tools that allow one to do tasks such as terminology creation before doing a machine translation. Although I agree that this is a very good way to improve quality, I wonder if you have any resources about using Google Translate (i.e. a machine translation system that does not permit the user to manipulate its translations before suggesting them). For freelancers in the less common languages, free online machine
... See more
@Jeff:

Most of your articles that I've seen relate to using commercial tools that allow one to do tasks such as terminology creation before doing a machine translation. Although I agree that this is a very good way to improve quality, I wonder if you have any resources about using Google Translate (i.e. a machine translation system that does not permit the user to manipulate its translations before suggesting them). For freelancers in the less common languages, free online machine translation will probably remain the only machine translation they have access to for some time to come, don't you agree? In my own language combinations we have only Google and Apertium (and only Google is currently of use to professional translators, I think).
Collapse


 

Jeff Allen  Identity Verified
France
Local time: 19:53
Multiplelanguages
+ ...
can you customize GT and other online MT portals for less prevalent languages Sep 1, 2009

Samuel Murray wrote:

@Jeff:

Most of your articles that I've seen relate to using commercial tools that allow one to do tasks such as terminology creation before doing a machine translation. Although I agree that this is a very good way to improve quality, I wonder if you have any resources about using Google Translate (i.e. a machine translation system that does not permit the user to manipulate its translations before suggesting them). For freelancers in the less common languages, free online machine translation will probably remain the only machine translation they have access to for some time to come, don't you agree? In my own language combinations we have only Google and Apertium (and only Google is currently of use to professional translators, I think).


Thanks Samuel for a very pertinent question, mainly because whenever I hear someone say that they are using an online MT system for any of the main international languages (FIGES and a few others), I say "why use the lowest level, basic system when you can buy one that can let you customize the translated output?"
I focus on using the professional and expert versions of MT systems, but I also do have home and office level versions of MT sofware that provide some customization features, but usually only offer an basic or novice mode, rather than advanced mode.

That being said, you are taking a different angle, especially by the fact that Google Translate is now up to 51 languages. So, it's really the only MT resource available for many of the lesser prevalent languages.
Well, there is the possibility of taking an opensource system and training it with your own data, but that just not really possible to get anything usable.

I've explained in the 2 following posts how Statistics-based MT works (Google Translate being one of them) in not too technical terms, with an explanation of how it is analyzing and processing the data to get a relevant translation, as compared to how a TM tool works.

Google Translate is not doing TM segment matching like a TM tool
http://www.proz.com/forum/wordfast_support/133682-how_can_i_hook_up_google_translate_to_wordfast.html#1122434

statistical MT approach + TMs
http://www.proz.com/forum/translator_resources/100328-machine_translation:_your_experience_with_the_various_mt_programmes_state_of_play-page2.html#998639

The hardest thing is getting enough data to use, and especially for your language pair. Google has access to enormous amounts of text on the web, so they can experiment in all kinds of ways.
And since a Statistical based system is giving probablistic guesses, there is very little control of it for these free online systems (this is not true of the customizable, commercial Statistical MT systems). With a free system like GT, you have no way to feed anything into it. The same is true for the online free rule-based MT engines. You get what you get, except possibly for SystraNet which I believe does have some customization options.

So for the less prevalent languages, there is not much you can do but use the baseline system that is trolling for usable training wherever it can.

However, the landscape may change in the coming years as some of the commercial Statistical MT vendors have said they are investigating ways to provide their MT systems in a SAAS (software as a service) model. And some of them are moving into various less prevalent languages, so maybe they will start offering such systems on a more widespread level.

As has been discussed at a few round tables about MT for minority languages over the years, MT vendors always say that the upfront investment cost for making a new language pair from scratch is several hundred thousand dollars/euros. Funding is necessary, and that's why you don't see as many languages offered by the mainline MT vendors. Somebody has to pay for that research and development work, and they won't do it themselves if they don't see the financial payback which can keep their company surviving. There are various posts out there that give the total amount of revenue for all translation tool vendors worldwide, and its not huge when you consider it is the consolidated amount of revenue for all of them.

So until someone wants to fund such projects (and there is usually some kind of motivation behind that), there is not much you can do to improve the translation of such free online systems. They keep adding training data over time, so that will increase the translation quality progressively, but not in a radical way that you can do with dictionary modules in paid commercial MT software.

There are some academic MT systems being created here and there, and I know that there has been a lot of interest in over the years for MT systems for languages in South Africa, because there are several language there which have official status. And such official status often generates a need for translation.

There is have the Google Translation Kit which seems to be moving in the direction of mixing your own TMs with the Google Translate MT systems. However, as has already been mentioned in other posts on that topic, Google reserves the right to reuse your content at their discretion, so that causes problems for the freelance translation community. Such an online system moves in the direction of online customization based on your own content, but it brings with it the disadvantage of having to cede the use of that content to the tool provider.

So, sorry that there is not much more advice I can provide on this specific question. Time will tell with the improvement of translation quality, with the available of features, with the availability of new language pairs, and with the modes of making systems available to the professional translator community. Google doesn't see professional translators yet as a customer, and so they are relying on a much wider general public to provide the upload of content, whereas a translator would be quite reticent about uploading their precious TM to a TM tool to improve MT.

Maybe they will introduce a terminology tweaking application over time. but it's not available at this time.

Jeff


 

Jeff Allen  Identity Verified
France
Local time: 19:53
Multiplelanguages
+ ...
info on Eng>Hindi MT systems Sep 1, 2009

Sushan Harshe wrote:

Post title: MT's are still in beginning stage

This is just an aggressive marketing strategy, below is an example of MT En>Hi
"Come home to take your food = आओ घर अपना खाना लेने के लिए"
This is very poooore literal translation. It has to walk, toooo long way to achieve at least understandable quality of language.


Jeff Allen wrote:
Please be more explicit and show more rigour in info provided before making general blanket statements about the status of MT as a whole

EN > HI

The language direction is specified. That is a good start.
Now:
Which system? (Commercial, industry research system, academic research system?) Which version number of the system? (for an online system you usually won't get the version number, but the date tested would be minimal info)
....



I had forgotten about the following post that I made a couple of years ago on the status of availability of HI MT systems.

Eng>Hindi MT
http://www.proz.com/forum/translator_resources/17428-englishhindi_mt:_does_it_even_exist.html#174630

Jeff


 

Eric Hahn (X)  Identity Verified
France
Local time: 19:53
French to German
+ ...
This industrial revolution is already a reality Sep 2, 2009

"The calls are generally published in the Official Journal of the European Union, C series and may relate not just to translation proper but also to the rewriting, modification and revision of texts and the _rapid post-editing of machine translation output_."

Source : http://europa.eu/languages/en/chapter/35


The only question is : What will be the effects of this
... See more
"The calls are generally published in the Official Journal of the European Union, C series and may relate not just to translation proper but also to the rewriting, modification and revision of texts and the _rapid post-editing of machine translation output_."

Source : http://europa.eu/languages/en/chapter/35


The only question is : What will be the effects of this industial revolution ?

Will the automatization lead to a further mechanization of translation work?

Will translation work become as rewarding as assembly line work ?

[Modifié le 2009-09-02 10:46 GMT]

[Subject edited by staff or moderator 2009-09-03 11:07 GMT]
Collapse


 

Samuel Murray  Identity Verified
Netherlands
Local time: 19:53
Member (2006)
English to Afrikaans
+ ...
SITE LOCALIZER
Thanks for your insights, Jeff Sep 2, 2009

Jeff Allen wrote:
...whenever I hear someone say that they are using an online MT system for any of the main international languages (FIGES and a few others), I say "why use the lowest level, basic system when you can buy one that can let you customize the translated output?"


It may have something to do with being willing to make an investment. There are many cases where one can buy a better product than the free one, but one must be willing to change one's ways drastically to get the full benefit from the purchase. For example, you can buy TO3000 for improved invoicing, but if you don't change your entire book-keeping system, you might as well have stayed with a free spreadsheet. Ditto buying an MT system -- to make the purchase worth it, you need to change the way you translate, the way you think and the way you type. If you use a free service, you can decide for yourself how much you'd like to get out of it.

...but I also do have home and office level versions of MT sofware that provide some customization features, but usually only offer an basic or novice mode, rather than advanced mode.


I think another reason why translators might be less inclined to buy an MT system is the onus of responsibility for compatibility with their CAT tool. If my CAT tool claims to work with Google Translate, and something doesn't work, I can turn to the CAT tool vendor and expect him to sort it out. But if I use a commercial MT system, and my CAT tool fails to connect to it, my CAT tool vendor (who does not own a copy of each and every MT system on the market) might have cause to say "well, ask the MT vendor to sort out the problem" (and the MT vendor will say a similar thing about the CAT tool vendor). So in a sense buying an MT system is a risk (although the rewards are typically higher as well).

Well, there is the possibility of taking an opensource system and training it with your own data, but that just not really possible to get anything usable.


Do you happen to know how much data is necessary to train an opensource system up to a useable level (assuming we can define "usable level" easily)? I realise it is a silly question (with relativity and all that) but I'm hoping for a vague indication here.

With a free system like GT, you have no way to feed anything into it.


Yes. I think the best way to provide useful information for freelancers using GT therefore is to determine what kinds of problems they're likely to encounter (and how to solve it/avoid it), what kinds of things they can expect to get from GT (or: in what ways can GT be useful for the translator), and just generally information about what a translator can expect if he has never used GT for a professional job before.

So until someone wants to fund such projects (and there is usually some kind of motivation behind that), there is not much you can do to improve the translation of such free online systems.


The answer may lie in crowdsourcing. Google is already doing it by giving users the opportunity to suggest a better translation. I'm not sure what Google does with such submitted translations, though.

...I know that there has been a lot of interest in over the years for MT systems for languages in South Africa, because there are several language there which have official status. And such official status often generates a need for translation.


I'm not aware of any MT systems being developed for (non-Germanic) South African languages. I've heard rumours of an English-Afrikaans engine developed by some university about ten years ago, but although I encounter the rumour time and again, I've been unable to confirm it myself.

The non-Germanic languages in South Africa have little economic power, and their speakers are often ashamed of using them. If you want to sell a product to an Afrikaans person, write your advert in Afrikaans (because Afrikaners are proud and they have money). But if you want to sell the same product to a Sotho or Venda or Tshwana person, don't use the person's own language -- he might think your product is weird.

There is have the Google Translation Kit which seems to be moving in the direction of mixing your own TMs with the Google Translate MT systems.


From my own limited experience with mixing MT with TM I can only say that editing a TM match and editing an MT match are two completely different types of tasks.

Time will tell with the improvement of translation quality, with the available of features, with the availability of new language pairs, and with the modes of making systems available to the professional translator community.


Yes, I think the focus of freelance translators using GT should for now be on improved workflow and not on getting an improved system (which, after all, is closed and therefore unimproveable).


 

Didier Briel  Identity Verified
France
Local time: 19:53
English to French
+ ...
MT for (non-Germanic) South African languages Sep 2, 2009

Samuel Murray wrote:
I'm not aware of any MT systems being developed for (non-Germanic) South African languages. I've heard rumours of an English-Afrikaans engine developed by some university about ten years ago, but although I encounter the rumour time and again, I've been unable to confirm it myself.


Have you checked Autshumato ITE?

It is partly based on OmegaT, but includes also machine translation. According to the documentation, it covers:
English, en_ZA
Afrikaans, af_ZA
IsiNdebele, nr_ZA
IsiZulu, zu_ZA
IsiXhosa, xh_ZA
Sesotho, st_ZA
Siswati, ss_ZA
Setswana, tn_ZA
Sepedi, sk_ZA
Tshivenda, ve_ZA
Xitsonga, ts_ZA

I have now idea about the level of machine translation actually provided, and between which pairs.

Didier


 

Samuel Murray  Identity Verified
Netherlands
Local time: 19:53
Member (2006)
English to Afrikaans
+ ...
SITE LOCALIZER
MT for 11 languages packed into 7 megabytes Sep 2, 2009

Didier Briel wrote:
Have you checked Autshumato ITE? ... It is partly based on OmegaT, but includes also machine translation. According to the documentation, it covers...


Well, you can think for yourself... machine translation between 11 languages, packed into 7 megabytes?

No, it doesn't "cover" those language -- it simply does not give the user the option to choose any other languages as his source and target language when translating a document (it is, after all, a product meant exclusively for the South African languages).

The user manual does not mention machine translation at all -- I suspect the program will eventually connect to some server... in version 1.0 or later.

BTW, Autshumato is being developed by people from the North West University in South Africa -- AFAIK they don't have an MT design programme although they do develop spell-checkers for South African languages. They should have access to a lot of multilingual content, though.


[Edited at 2009-09-02 18:52 GMT]


 

Didier Briel  Identity Verified
France
Local time: 19:53
English to French
+ ...
Not everything is packaged Sep 2, 2009

Samuel Murray wrote:

Didier Briel wrote:
Have you checked Autshumato ITE? ... It is partly based on OmegaT, but includes also machine translation. According to the documentation, it covers...


Well, you can think for yourself... machine translation between 11 languages, packed into 7 megabytes?


I didn't say everything was packaged, ready to use. By "check", I mean try and get information.


No, it doesn't "cover" those language -- it simply does not give the user the option to choose any other languages as his source and target language when translating a document (it is, after all, a product meant exclusively for the South African languages).

The user manual does not mention machine translation at all -- I suspect the program will eventually connect to some server... in version 1.0 or later.

It already does, although rather experimentally I guess. See a message from the developer.

Unless things have changed, they plan to use Moses for the MT engine.

Didier


 
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?

Advanced search






WordFinder Unlimited
For clarity and excellence

WordFinder is the leading dictionary service that gives you the words you want anywhere, anytime. Access 260+ dictionaries from the world's leading dictionary publishers in virtually any device. Find the right word anywhere, anytime - online or offline.

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

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search