Practical advice for MT translators
Thread poster: Samuel Murray

Samuel Murray  Identity Verified
Netherlands
Local time: 20:37
Member (2006)
English to Afrikaans
+ ...
Aug 27, 2009

G'day everyone

Google's MT engine finally does my language too. And I expect that the amount of work in my language in which Google's MT translations will play a role, will increase. This will include proofreading MT'ed translations or getting a TM from a client that had been MT'ed. Using MT is not new in other languages, though. So I'm wondering what resources you know of with some practical advice for translators moving into the MT-editing field, or who will be using MT in addition to TM (segment by segment) in their translations.

So my first question is about practical advice for professional translators. However, I just realised (when reading another ProZ.com post) that MT can also be useful in crowdsourcing (i.e. do crowdsourced MT proofreading). Do you know of any studies or resources about this?

Thanks
Samuel


 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 20:37
Member (2005)
English to Spanish
+ ...
Good to have at hand Aug 27, 2009

1st resource: Aspirin, lots of it!

2nd resource: PureCalm or equivalent, lots of it too!

You are bound to work as much as translating from scratch... but at 1/3 of the pay. That is bound to provide you with a fair deal of headaches and nervous breakdown if you consider MT editing a genuine option for your career.


 

Samuel Murray  Identity Verified
Netherlands
Local time: 20:37
Member (2006)
English to Afrikaans
+ ...
TOPIC STARTER
Not really useful advice Aug 27, 2009

Tomás Cano Binder, CT wrote:
You are bound to work as much as translating from scratch... but at 1/3 of the pay.


Not really useful advice.

I also doubt if I'll be translating from scratch. I just put a few Wikipedia pages through its paces and in many cases I only had to make two or three changes per sentence. I'm fortunate in that my two languages have rather similar word orders.


 

Sushan Harshe
India
Local time: 00:07
English to Hindi
+ ...
He is not wrong! Aug 27, 2009

Hi Samuel,

It is common experience, that M.T. has not reached to an extent where there could be very less to edit. You have said, "in many cases I only had to make two or three changes..."

how many trials you took with that MT. (and for how many times you may be using it in life!) My experience with MT is just same as Tomás Cano Binder (I do not know how to shorten his name) has advised you.

There are some euro languages which has similar grammar structure as English, MT works some what better with these language. (accuracy 60 to 70%) I do not know if the same applies to your language. For Indian languages MT accuracy is not more then 30%.

My observations (and answer to your questions) are:

1] creating job on MT is joyful.

2] editing in the name of review, of an MT generated job is very frustrating.

3] crowdsourcing is nothing but lucrative style of business. where language quality has a very less importance.

4] there are lots of forums and discussions about MT, on all over the web but, you may find hardly something knowledgeable (hence helpful). I did not found any study paper on web about MT technic.


 

Philippe Etienne  Identity Verified
Spain
Local time: 20:37
Member
English to French
Do not take MT "meaning" at face value Aug 27, 2009

Dear Samuel,

Some specific terminology is used, such as "MT post-editing", "raw output", "light post-editing", "full post-editing" or "controlled language" that may yield results on an Internet search.
Some members here are very knowledgeable about MT, I remember seeing links to resources on forums as well.

My piece of advice: it happens that MT generates the exact opposite of what is meant. You should bear this in mind when rereading in context even for sentences that "look" meaningful. This is especially true with some sentences containing negations.

Off topic to Tomas: you may regard MT as a specialised tool, like a jig saw: it does a pretty good job at cutting wood planks, but it is also well known for chopping fingers off. The main thing with tools is to know how and when to use them. Some texts (edit: and language pairs I guess) lend themselves pretty well to MT, a lot of others don't.

Philippe

[Edited at 2009-08-27 13:11 GMT]


 

Claudia Alvis  Identity Verified
Peru
Local time: 13:37
Member
Spanish
+ ...
Just say no Aug 27, 2009

MT is very easy to identify. You will end up re-translating the whole thing. So if you receive a request to "proofread" MT content, you could explain (or even show) your client that that it'd be more expensive for them-- you'd need to read the MT content, identify the differences between the meaning of original and the MT text, try to fix it, give up and re-translate the whole thing anyway.

I just say no. I had a few regular clients that started sending me MT to correct. So I sent them my old translations back-translated with MT, along with the originals. I didn't lost any of those clients.

About "controlled language", many of the large companies that are investing millions in MT research, and do use controlled language, also use human translators to translate that content.


 

Piotr Bienkowski  Identity Verified
Poland
Local time: 20:37
Member (2005)
English to Polish
+ ...
Exactly, watch out for negations Aug 27, 2009

Philippe Etienne wrote:

it happens that MT generates the exact opposite of what is meant. You should bear this in mind when rereading in context even for sentences that "look" meaningful. This is especially true with some sentences containing negations.


Exactly. I have seen that with google translate.

Piotr


 

Piotr Bienkowski  Identity Verified
Poland
Local time: 20:37
Member (2005)
English to Polish
+ ...
This is not always the case Aug 27, 2009

Claudia Alvis wrote:

MT is very easy to identify. You will end up re-translating the whole thing. So if you receive a request to "proofread" MT content, you could explain (or even show) your client that that it'd be more expensive for them-- you'd need to read the MT content, identify the differences between the meaning of original and the MT text, try to fix it, give up and re-translate the whole thing anyway.


Google's MT is pretty good with some types of documents, like legalese, everything EU related, and also some IT stuff, at least in my language pair.

Admittedly, with some other document/register types, its results are miserable.

Regards,

Piotr


 

FarkasAndras
Local time: 20:37
English to Hungarian
+ ...
EU Aug 27, 2009

Piotr Bienkowski wrote:

Google's MT is pretty good with ... everything EU related


Oh then I'm sure they just went ahead and fed in the Acquis corpus and the europarl corpus... Come to think of it, one could use it as a poor man's concordance search!


 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 20:37
Member (2005)
English to Spanish
+ ...
Completely agree Aug 28, 2009

Philippe Etienne wrote:
Off topic to Tomas: you may regard MT as a specialised tool, like a jig saw: it does a pretty good job at cutting wood planks, but it is also well known for chopping fingers off. The main thing with tools is to know how and when to use them. Some texts (edit: and language pairs I guess) lend themselves pretty well to MT, a lot of others don't.

Absolutely. In the case of English into Spanish, simple procedural texts require a relatively low amount of editing. MT is no help in my case however as it really takes knowledge and experience in the matter at hand, and so many cases also common sense when the writers had a bad day.icon_smile.gif


 

Marco Cevoli  Identity Verified
Spain
Local time: 20:37
Spanish to Italian
+ ...
A checklist of common pitfalls Aug 29, 2009

Hi Samuel,

I prefer not to enter in the discussion about the essence of MT. I'd rather give you my insights about the common pitfalls I face when I edit texts pretranslated with MT tools. This is, of course, limited to my 2 main working languages (EN>IT, ES>IT) and mainly to Google Translate, so please take it with a grain of salt.

- Inconsistencies: this is one of the most serious problems. Same terms are translated differently in different sentences (even though they come from the same MT batch).
- Acronyms: sometimes they are completely mistranslated or not translated at all.
- Capitalization: there is often a tendency of keeping as many capital letters as in the English source.
- Repetitions: depending on the sentence structure, mostly on long sentences, some words may appear twice (in different positions)
- Punctuation: dashes are often maintained even though in Italian we use colons instead. Other issues regarding punctuation also arise. Be careful.
- Names, titles and other proper nouns: sometimes they are literally translated, or worse, they are partially translated.
- Word order (very specific to EN>IT): MT almost always gives you Subject-Verb-Object order.
- Negations: I totally agree with the previous comments. See previous messages.

There are other general remarks, but I don't have them at hand right now.

Hope it helps

Best regards

Marco Cevoli
Qabiria

[Edited at 2009-08-29 21:28 GMT]


 


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


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

Practical advice for MT translators

Advanced search







CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use SDL Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

More info »
PerfectIt consistency checker
Faster Checking, Greater Accuracy

PerfectIt helps deliver error-free documents. It improves consistency, ensures quality and helps to enforce style guides. It’s a powerful tool for pro users, and comes with the assurance of a 30-day money back guarantee.

More info »



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