| Pages in topic: < [1 2 3 4 5 6 7 8 9 10] > | What's your opinion on machine translation and quality? Thread poster: Daniela Zambrini
|
|---|
| We all use it | May 23, 2014 |
No, I don't use MT to translate a whole text, indeed big laugh, irritation and extra work, but I find a lot of words on for example on "MyMemory", also MT, and I don't think I am the only one.
With or without realising it, we all embraced the future.
| | | | DZiW (X)
Ukraine
English to Russian
+ ...
| whole or in parts? | May 23, 2014 |
I didn't get whether we are talking about entire text translation or just about segments/sentences.
The former is for pre-provisional-alpha0.0.001a only, whereas the latter is quite acceptable sometimes.
For instance, sometimes I use MT for prompting (wording/meaning) mostly for short sentences, because out of context it usually goes wrong badly. Also I often find other translators/editors/proofreader using (checking?) several MT's to find out which one was actually used... Ma... See more I didn't get whether we are talking about entire text translation or just about segments/sentences.
The former is for pre-provisional-alpha0.0.001a only, whereas the latter is quite acceptable sometimes.
For instance, sometimes I use MT for prompting (wording/meaning) mostly for short sentences, because out of context it usually goes wrong badly. Also I often find other translators/editors/proofreader using (checking?) several MT's to find out which one was actually used... Makes sense
Shortly, if something is done with purpose and understanding (including risks assessment), then it's fine, but if it's just a lame excuse for laziness, then it's a big no-no: anything could be used for something good or something bad, unfortunately.
As for me, considering the environment, NDA, and other factors, I do refrain from using MT while translating. A CAT is enough, should do. ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 02:03
Member (2008)
Italian to English
| Useless even for simple things | May 23, 2014 |
Here's this post translated into Italian by Google Translate. It can't even get this right ! I rest my case.
"Ecco questo post tradotto in italiano da Google Translate. Non si può anche ottenere questo diritto! Io resto il mio caso."
The foregoing example demonstrates that MT is completely incapable of knowing the context in which something is said, or when a word such as "right" might mean either "correct" or "entitlement". Etc. Need I say more?
Nor does ... See more Here's this post translated into Italian by Google Translate. It can't even get this right ! I rest my case.
"Ecco questo post tradotto in italiano da Google Translate. Non si può anche ottenere questo diritto! Io resto il mio caso."
The foregoing example demonstrates that MT is completely incapable of knowing the context in which something is said, or when a word such as "right" might mean either "correct" or "entitlement". Etc. Need I say more?
Nor does it understand idiomatic English. For example:
"Labour took heart after it topped the local polls in England with 31% of the vote – up two points on last year."
"Labour ha preso il cuore dopo aver superato i sondaggi locali in Inghilterra con il 31% dei voti - due punti dello scorso anno."
Completely wrong !
[Edited at 2014-05-23 21:02 GMT] ▲ Collapse
| | | | | Post removed: This post was hidden by a moderator or staff member for the following reason: empty post
|
|
|
| If I want quality I dont use it | May 24, 2014 |
I used to be asked to post-edit a machine-translated text and it was all wrong. I had to erase the whole thing and create it from scratch. It is just a series of mismatched words, not a meaningful sentences. Some time it reversed the whole meaning like saying what you shouldn't do into what you should do.
Lately, I've been asked to review a translation for an agency which was submitted to client and after one year they came back with questions on the quality. I looked at it and it t... See more I used to be asked to post-edit a machine-translated text and it was all wrong. I had to erase the whole thing and create it from scratch. It is just a series of mismatched words, not a meaningful sentences. Some time it reversed the whole meaning like saying what you shouldn't do into what you should do.
Lately, I've been asked to review a translation for an agency which was submitted to client and after one year they came back with questions on the quality. I looked at it and it turns out to be a post-edited machine translation. It seemed the post-editor wasn't paid well enough to notice and clear all its errors and some of the target text still have source text in it (not retained words that were specified or should be retained).
So it is more productive and beneficial to not use MT. What we need is built glossary and terms. ▲ Collapse
| | | | | I know it is still struggling in my language pairs, but making some progress | May 24, 2014 |
With controlled language and dedicated search engines it can produce results that give people an idea of what a text is about. Some sentences come out intelligibly, while in others the source language syntax is difficult to read, although not strictly wrong in English. Others are not translated or simply wrong.
I do not like post-editing - I find I soon get 'blocked' and would prefer to translate from scratch.
It has trouble with verb congruence and a lot of other issue... See more With controlled language and dedicated search engines it can produce results that give people an idea of what a text is about. Some sentences come out intelligibly, while in others the source language syntax is difficult to read, although not strictly wrong in English. Others are not translated or simply wrong.
I do not like post-editing - I find I soon get 'blocked' and would prefer to translate from scratch.
It has trouble with verb congruence and a lot of other issues, and I have signed agreement with some of my clients that I will not use it, though largely for security reasons.
One of these clients is working on it in a separate department from the translation section, but that might explain why they do not want their translators using it (Not yet at any rate!)
I prefer to have nothing to do with it. ▲ Collapse
| | | | | Yes absolutely. | May 24, 2014 |
There is no such a things as 'grammatical mistakes to be corrected". If the grammar is wrong --everything is wrong, because the text does not mean whatever it was supposed to mean. The people who can really assess the quality of translation know very well that MT, unfortunately, has no future in professional translation because it lacks the intelligent agent--a human being. Those who can't, and believe in advertising, think is is a wonder of the 21st century. MT may have some other uses, perhaps... See more There is no such a things as 'grammatical mistakes to be corrected". If the grammar is wrong --everything is wrong, because the text does not mean whatever it was supposed to mean. The people who can really assess the quality of translation know very well that MT, unfortunately, has no future in professional translation because it lacks the intelligent agent--a human being. Those who can't, and believe in advertising, think is is a wonder of the 21st century. MT may have some other uses, perhaps--to a limited degree: in language teaching and entertainment.
[Edited at 2014-05-24 11:15 GMT] ▲ Collapse
| | | | Joakim Braun 
Sweden
Local time: 03:03
German to Swedish
+ ...
It's progress: you can now get a rotten translation for nothing, instead of having to pay someone $0.00001 per word for it.
| | |
|
|
|
neilmac
Spain
Local time: 03:03
Spanish to English
+ ...
In principle I shoudl be anti-MT, but I'm afraid that in my case I only disapprove of its use when the results are sub-standard. I actually use MT as part of my arsenal as a translator and have done so for about 15 years now. If used judiciously, it can save wear and tear on your wrist and back, etc and replace a lot of typing, in a similar way to cut and paste. Initially I found that an early version of Systran was useful to help with certain types of texts - mainly official documents with a bu... See more In principle I shoudl be anti-MT, but I'm afraid that in my case I only disapprove of its use when the results are sub-standard. I actually use MT as part of my arsenal as a translator and have done so for about 15 years now. If used judiciously, it can save wear and tear on your wrist and back, etc and replace a lot of typing, in a similar way to cut and paste. Initially I found that an early version of Systran was useful to help with certain types of texts - mainly official documents with a bureaucratic style. However, when it was upgraded for use with 64-bit operating systems, I found the new versions no longer user-friendly and gave up on it. Nowadays I used GT4T instead and although it works by a different system, I find I get similar results to when I was using Systran. I occasionally find it useful to compare a sentence, or chunk, produced by MTV with my own efforts - it's sort of like having a colleague or friend nearby to bounce ideas off when there is no one around. However, I don't use it to translate whole paragraphs or pages, only small chunks of text or occasionally a full sentence.
I also started using Dragon naturally speaking a few years ago, attempting to take some of the strain off my typing arms and, as with MT, it entails slightly more painstaking revision than when I type, but I now find I can translate more quickly using a mixture of Dragon, MT and typing... And I can go for longer without becoming physically exhausted.
Perhaps more worrying is the fact that many companies, governments or similar institutions no longer bother translating their material or websites other than by machine translation. Yesterday I access a Spanish government site which offered a machine translated English version (far from perfect, but more or less understandable) accompanied by a caveat/disclaimer along the lines of "this page has been translated automatically without review by a translator....". Unfortunately, it looks like this is one genie we won't be able to put back in the bottle...
[Edited at 2014-05-25 08:37 GMT] ▲ Collapse
| | | | | You can still stop it form spreading | May 25, 2014 |
neilmac wrote:
Perhaps more worrying is the fact that many companies, governments or similar institutions no longer bother translating their material or websites other than by machine translation. Yesterday I access a Spanish government site which offered a machine translated English version (far from perfect, but more or less understandable) accompanied by a caveat/disclaimer along the lines of "this page has been translated automatically without review by a translator....". Unfortunately, it looks like this is one genie we won't be able to put back in the bottle...
[Edited at 2014-05-25 08:37 GMT]
Into important areas of life--like science, law, medicine, politics and art.
| | | | | I think we have to accept it | May 25, 2014 |
It IS making progress, and for routine translations, WITH the usual caveats and corrections, it may be acceptable.
There are simply not enough humans to do all the necessary translations, and in its place I prefer MT to crowdsourcing, which really does degrade the profession and give people the idea that any child can do it!
You can programme a machine to recognise the standard remarks and reporoduce them with similar standard remarks fairly reliably.
Then ... See more It IS making progress, and for routine translations, WITH the usual caveats and corrections, it may be acceptable.
There are simply not enough humans to do all the necessary translations, and in its place I prefer MT to crowdsourcing, which really does degrade the profession and give people the idea that any child can do it!
You can programme a machine to recognise the standard remarks and reporoduce them with similar standard remarks fairly reliably.
Then we can concentrate on the areas where humans are really needed. ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 02:03
Member (2008)
Italian to English
Christine Andersen wrote:
It IS making progress, and for routine translations, WITH the usual caveats and corrections, it may be acceptable.
There are simply not enough humans to do all the necessary translations, and in its place I prefer MT to crowdsourcing, which really does degrade the profession and give people the idea that any child can do it!
You can programme a machine to recognise the standard remarks and reporoduce them with similar standard remarks fairly reliably.
Then we can concentrate on the areas where humans are really needed.
I can't agree. I've been watching MT for some years now and I don't see it making any progress at all. It still cannot understand nuance and idiom, and cannot even translate a simple statement without making a complete dog's breakfast of it.
For example GT translates "cannot even translate a simple statement without making a complete dog's breakfast of it" into Italian as "non può nemmeno tradurre una semplice dichiarazione senza fare la prima colazione di un cane completa di esso"
... and this is ordinary, everyday English. You'd expect GT to at least get that right, but it can't.
I suppose MT may be useful - but not for anything important, because the mistakes it makes are very serious.
[Edited at 2014-05-25 10:49 GMT]
| | |
|
|
|
| I agree with Tom--No Good at All | May 25, 2014 |
Maybe you have just experience with Scandinavian-English translations (Norwegian, Danish, Swedish) into English, Christine. I must admit that in those pairs it is not horrible due to the similarity of the languages on a deeper level, plus in my experience, people in Scandinavia do not use too many redundant words and very flowery structures, which may also be a factor. I would not rely on it, though, for any serious purpose.
Wait till you see some Slavic or Baltic languages--it is... See more Maybe you have just experience with Scandinavian-English translations (Norwegian, Danish, Swedish) into English, Christine. I must admit that in those pairs it is not horrible due to the similarity of the languages on a deeper level, plus in my experience, people in Scandinavia do not use too many redundant words and very flowery structures, which may also be a factor. I would not rely on it, though, for any serious purpose.
Wait till you see some Slavic or Baltic languages--it is even too scary to laugh at some of the translations. Just horror. Some words are even mixed between different languages.
I think MT may have some use for set phrases translation (idioms, certain terms in various areas of life--country-specific), but not any texts, or even sentences, in their entirety.
[Edited at 2014-05-25 11:36 GMT] ▲ Collapse
| | | | | I agree, GT is probably never going to work | May 25, 2014 |
And the websites one sees translated with it are a mess.
There may be possibilities using controlled language - which I worked on monolingually way back in the 1970s, computers can recognise set phrases. (At that time it was for searching library indexes and abstracts for relevant articles and documents, and it was in fact quite useful even then.)
Here pre-editing was an important part of the process, and with some kinds of standardised 'texts' and modern computers with... See more And the websites one sees translated with it are a mess.
There may be possibilities using controlled language - which I worked on monolingually way back in the 1970s, computers can recognise set phrases. (At that time it was for searching library indexes and abstracts for relevant articles and documents, and it was in fact quite useful even then.)
Here pre-editing was an important part of the process, and with some kinds of standardised 'texts' and modern computers with fuzzy logic, it is possible to match up corresponding prepared controlled phrases in the target language. (What I call the CAT principle, but I had better not pick an argument with you about CATs!)
It helps a lot if the subject area is retricted too - to a single specific field, where the homonyms are limited.
Otherwise I agree - I have seen some ridiculous renderings of 'business executives wearing shorts for security reasons (Stylish Swedes in Stockholm!!) and no, they were carrying ID cards. (Homonym-same verb for carrying and wearing, and the word for card is a homonym of the adjective short in Swedish...)
Or gibberish about the Act on Dividing a Court - in a review of a restaurant (The source meant 'We were allowed to share a Starter = First Course.)
One researcher explained that MT has a lot of trouble with compound verbs, and gave quite a funny example, which unfortunately is untranslatable, about a young man breaking up with his girlfriend, though not because he beat her
(slog op med hende, but not slog hende).
I agree that there will be a lot of areas where translation is way beyond computers for the foreseeable future.
And Lilian, relaying through intermediate languages is definitely asking for trouble - that can only be done by humans who know the subject area, or else you will get results with bits in wrong languages like GT's efforts with the Baltic languages.
The whole process has to be repeated from scratch for each language pair, so there is still a long way to go with many languages.
As I said, I don't have much confidence in MT, but never say never!
[Edited at 2014-05-25 19:38 GMT] ▲ Collapse
| | | | Oleg Vigodsky
Russian Federation
Local time: 05:03
English to Russian
LilianBNekipelo wrote:
Maybe you have just experience with Scandinavian-English translations (Norwegian, Danish, Swedish) into English, Christine. I must admit that in those pairs it is not horrible due to the similarity of the languages on a deeper level, plus in my experience, people in Scandinavia do not use too many redundant words and very flowery structures, which may also be a factor. I would not rely on it, though, for any serious purpose.
Wait till you see some Slavic or Baltic languages--it is even too scary to laugh at some of the translations. Just horror. Some words are even mixed between different languages.
I think MT may have some use for set phrases translation (idioms, certain terms in various areas of life--country-specific), but not any texts, or even sentences, in their entirety.
[Edited at 2014-05-25 11:36 GMT]
Dear collegues,
I understand you have almost never seen good MT. Me either . Except for some good examples and 'long-term' solutions. Bad MT output does not mean bad technology, but (in most cases) bad MT customization and/or training.
If a use a 3rd party TM (suppose, TRADOS) and I see a chaos with terms and consistency, I do not blame a SW, but people responsible for a terminology development, translation proof-reading and TM administration (if any).
The same case is with MT. Very few companies (LSPs and final Customers) are able to provide and manage a good MT output (in my experience). I wish you will get proper customized MT output. It really simplifies a translator's work.
PS: I use MT tools since 1993. And prefer to customize an MT system myself (with a feedback from my collegues).
| | | | | Pages in topic: < [1 2 3 4 5 6 7 8 9 10] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » What's your opinion on machine translation and quality? | 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 Trados, MemoQ, Wordfast & major CAT tools.
Download and start using CafeTran Espresso -- for free
Buy now! » |
| | 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 
