Getting into MT for massive translation turnarounds
Thread poster: Diego Achío

Diego Achío  Identity Verified
Local time: 01:09
Member (2010)
English to Spanish
+ ...
Jul 16, 2015

I've been reading a lot from sites like TAUS, SYSTRAN, and such. They all claim that Translation Agencies leading the translation world now use Machine Translations for projects above 10,000 words. They just outsource post-editing services from Freelancers, then use MT evaluation tools to assest the quality of the translation and apparently that saves them a lot of money.

I've always have a bad impression of MT, I've tested Microsoft's translation API which is supposed to be a Corporate application, Trados BeGlobal, and other resources but (of course) the more technical the content gets, the more annoying the machined translation becomes.

I know certain projects are very ambitious, such as videogame localization. A videogame may easily reach 300,000 words and the videogame developers require these translations as soon as possible. But I've always thought (silly me?) that humans were the ones translating.

In my experience (still talking about videogames), I've been hired to translate videogames with 50,000 words. So, what is the truth behind MT?


 

Soonthon LUPKITARO(Ph.D.)  Identity Verified
Thailand
Local time: 13:09
Member (2004)
English to Thai
+ ...
Video game Jul 17, 2015

Diego Achío wrote:

In my experience (still talking about videogames), I've been hired to translate videogames with 50,000 words. So, what is the truth behind MT?


Video game is similar to literature. Translation needs idiomatic review and MT is poor on this aspect. Human interference is more essential than MT on technical, medical or legal text translation.

Soonthon L.icon_biggrin.gif


 

Ahmed ALEM
Algeria
Local time: 07:09
Member (2014)
Arabic to English
+ ...
The concept of MT Jul 17, 2015

Hello,
Machine translation is the most controversal topic today when speaking about Translation, especially in the presence of such growing demand on Translation services in a very limited time, the thing that led Translation industry take new dimensions. Machine Translation aims at helping in the translation process and here many people were confused ! Do these softwares translate or help translate? Well in my attempt to ask this question i would say both answers since there are two types ofTranslation softwares;
a) Translation softwares: which comes befroe CAT tools, these are softwares that translate directly either word for word like dictionaries or texts like Golden Al-Wafi for Translation for example.
b) CAT tools: which stand for Computer Assisted Translation Tools; these tools do not translate automatically like the previous softwares that I mentioned, the thing is that developpers did not complete the task of forming a perfect translation softwares, they programmed them from all the technical aspects and let the linguistic part to fill by the translator, and that is how Translation Memories came to life in this field, so you create a TM and import to it all your previous translations and later you can use them on your future projects soif the softwares detects in the text similar segments that are already translated in those Translation Memories, the softwares here would just translate automatically basing on the Translation memories that you have already given to it.

So in order to guarantee a good marketing for these CAT tools, it is said that they save time, and yes they do save time alot when knowing how to use them perfectly, but this might be a negative point, sometime you face some outsourcers that would just require a very fast translation because they assume that the translator that they hire can do it fastly thanks to these CAT tools, I add here that such outsourcers have nothing to do with the field of Translation and have no idea how Translation process work.

So Machine Translation concept still remains mysterious and we will never know what we will see in few years later.

I hope y answer helped you.

[Edited at 2015-07-17 01:43 GMT]

[Edited at 2015-07-17 01:52 GMT]


 

Diego Achío  Identity Verified
Local time: 01:09
Member (2010)
English to Spanish
+ ...
TOPIC STARTER
Thanks but... Jul 17, 2015

alemahmed wrote:

Hello,
Machine translation is the most controversal topic today when speaking about Translation, especially in the presence of such growing demand on Translation services in a very limited time, the thing that led Translation industry take new dimensions. Machine Translation aims at helping in the translation process and here many people were confused ! Do these softwares translate or help translate? Well in my attempt to ask this question i would say both answers since there are two types ofTranslation softwares;
a) Translation softwares: which comes befroe CAT tools, these are softwares that translate directly either word for word like dictionaries or texts like Golden Al-Wafi for Translation for example.
b) CAT tools: which stand for Computer Assisted Translation Tools; these tools do not translate automatically like the previous softwares that I mentioned, the thing is that developpers did not complete the task of forming a perfect translation softwares, they programmed them from all the technical aspects and let the linguistic part to fill by the translator, and that is how Translation Memories came to life in this field, so you create a TM and import to it all your previous translations and later you can use them on your future projects soif the softwares detects in the text similar segments that are already translated in those Translation Memories, the softwares here would just translate automatically basing on the Translation memories that you have already given to it.

So in order to guarantee a good marketing for these CAT tools, it is said that they save time, and yes they do save time alot when knowing how to use them perfectly, but this might be a negative point, sometime you face some outsourcers that would just require a very fast translation because they assume that the translator that they hire can do it fastly thanks to these CAT tools, I add here that such outsourcers have nothing to do with the field of Translation and have no idea how Translation process work.

So Machine Translation concept still remains mysterious and we will never know what we will might see in few years later.

I hope y answer helped you.

[Edited at 2015-07-17 01:43 GMT]


Thanks alemahmed. I know what CAT tools are, I actually use MemoQ for almost all my projects. When I talk about MT I don't refer to Google Translator or Bing Translator. I mean today's sofisticate API options for which you usually have to pay per quantity of words translated or they charge a yearly membership.

Examples:

https://www.microsoft.com/translator/api.aspx
http://www.translationsoftware4u.com/api.php

[Editado a las 2015-07-17 01:52 GMT]

Actually, these translation APIs can be implemented into CAT tools so they "help you" translate. I tried implementing Microsoft Translation API and it is far more accurate than online translators. But they still fail by a lot when they face technical texts.

So, I'm wondering, is investing into a MT really worth it? I've seen companies that offer over 500,000 words delivered in impossible human turnarounds, and they even have quality ISO certifications! How in the world they accomplish that?!

[Editado a las 2015-07-17 01:56 GMT]


 

Steven Segaert
Estonia
Local time: 09:09
Member (2012)
English to Dutch
+ ...
Quality Jul 17, 2015

"How in the world they accomplish that?!"

Varying quality.

I have experimented with purpose-developed machine translation in a specialised area where it was possible to create clear rules, at least for sub-segments. In my experience, the results are disappointing for two reasons.

First, the unequal quality of source materials. As humans, we read what the writer means. We have no trouble interpreting typo's, missing punctuation, wrong expressions, bad line breaks... Unless text is specifically prepared and formatted with machine translation in mind, this is a first very big hurdle.

Second, the word-for-word approach that machine translation necessarily applies. If a certain sentence or word group is not in a memory, the machine can only translate like a dictionary. It can't get creative with word order and segmentation, and it has trouble with composite words and verbs (such as, for instance, in Dutch).

Even the best machine translation reads like a translation. The longer the phrases to be translated, the more that becomes evident.

Machine translation is a good solution for certain types of text, and certainly a good addition to the toolbox of making more materials accessible. But it is, in my opinion, no substitute for good human translation.

Clients looking to make repeating materials accessible in a cheap way should put time and effort in structuring their source materials accordingly and in developing good and unambiguous glossaries and style rules. With these in place, machine translation can work. In that respect, "pre-MT editing" makes a lot more sense.

Clients who want their materials to read like it was written by a human, should turn to a human. Even those who do engage in post-editing admit that the end result is perhaps correct, but not as good as a good human translation.

But once you understand what machine translation is capable of, and what it can't do, the idea of post-editing a machine result is rather silly. If the machine result is not good enough after some very marginal editing, it won't be good enough after heavy editing. Simply because the machine doesn't perform, or the materials are not good for machine translation.

If the machine result IS decent and produces good parts of sentences, the effort needed to craft good text out of these results is only marginally less than the effort needed to translate from scratch. You need to read and understand the source, read and understand the machine result, and then create a correct edit. It is doable, but it isn't easy. And after very careful consideration, I have decided for myself that it doesn't allow me to work as fast as the offered rates would suggest.

It's also a myth that a machine gets better by post-editing. It doesn't, if the rules it works with stay the same. At best, you just get a big translation memory of questionable quality.

Machine translation certainly has its place. Right now, there are just an awful lot of people deceiving others about what exactly that place is.


 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Getting into MT for massive translation turnarounds

Advanced search







memoQ translator pro
Kilgray's memoQ is the world's fastest developing integrated localization & translation environment rendering you more productive and efficient.

With our advanced file filters, unlimited language and advanced file support, memoQ translator pro has been designed for translators and reviewers who work on their own, with other translators or in team-based translation projects.

More info »
Protemos translation business management system
Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!

The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.

More info »



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