Sharing opinion about machine translation
Thread poster: Lexcode_kr

Lexcode_kr
South Korea
Mar 8

Hi,

I’ve posted it because I’d like to listen to translator’s thought about machine translation.
Many companies have launched services that are connected with machine translation or tried to catch up with that trend.
I have curiosity on what translators think about that.

So, do you use the machine translation when you translate? If you do, how frequently? And what kind of it? And what do you think its merit?
And the result was as accurate as you can use that directly?

If you have negative opinion on that, please let me know your opinion freely.

Thank you.


 

Heinrich Pesch  Identity Verified
Finland
Local time: 03:15
Member (2003)
Finnish to German
+ ...
Useful tool Mar 10

It seems nobody has hat answered your posting after 2 days.
I have GT always activated and get from it useful hints about terminology. Its far from perfect but better than a dictionary, as it keeps up with time. I have noticed many times when translating that when the same term pops up time and time again MT can suggest surprising alternatives for the translation, and I had often to admit that one suggestions is better than my initial selection.
The database of MT is far larger than any translator's translation memory and consists of thousands of translations done by different people.
I still have to open my paper dictionaries though a few times during a larger project, and of course the normal research as usual.
One useful feature of MT is that if it comes up with a target word this word is correctly spelled, even if it is the wrong word or needs adaptation to the grammatical structure.


 

Hans Lenting  Identity Verified
Netherlands
Member (2006)
German to Dutch
Agree and some nuances Mar 10

Heinrich Pesch wrote:

I have GT always activated and get from it useful hints about terminology. Its far from perfect but better than a dictionary, as it keeps up with time.


I agree. Especially for technical translations, these MT systems can be very useful, since they often harvest from the websites (with technical documentation) from the machine and tool manufacturers that are named in the technical documentation that I translate (or from their competitors).

I have noticed many times when translating that when the same term pops up time and time again MT can suggest surprising alternatives for the translation, and I had often to admit that one suggestions is better than my initial selection.


This is indeed the case. But this advantage can also become a disadvantage: one has to make sure that consistency (of part names etc.) is maintained. Good CAT tools offer features for this (by improving MT suggestions with your own terms).

MT systems crawl the web and harvest from translated sites (and documentation). I harvest from them: whenever I see a nice translation suggested in the 4 MT systems that I'm currently using parallel, I add these nice alternatives to my glossary. What I cannot use today, can be very useful tomorrow. (Surely, for this you have to have a CAT tool that makes the entry of alternative target terms a jiffy.)

The database of MT is far larger than any translator's translation memory and consists of thousands of translations done by different people.


You can indeed learn a lot from the hard work of your colleagues. It's a nice intellectual activity. Sometimes I hear myself saying: 'Wow ...'.

I still have to open my paper dictionaries though a few times during a larger project, and of course the normal research as usual.


On a side note: you don't use the electronic versions of these dictionaries?

One useful feature of MT is that if it comes up with a target word this word is correctly spelled, even if it is the wrong word or needs adaptation to the grammatical structure.


I cannot confirm that MT suggestions are always spelled correctly. Even good systems like DeepL create bogus words (by glueing words together). It would be nice to have an option to prevent word creation (word invention).

Related to this: I find the MT systems to provide a nice typing aid, since many technical documents have a lot of typos. Even these typos (words in the source text) are often recognised correctly by the MT systems. A great timesaver.


 

neilmac  Identity Verified
Spain
Local time: 02:15
Spanish to English
+ ...
Agree with Heinrich Mar 10

I use GT4T and it's helpful and sometimes amusing, but never a panacea. You still need to use other resources, such as your own brain/memory. I tend to only use it on short chunks of text, or clauses, sometimes just to see what it suggests and as Heinrich mentioned, sometimes they are better than my own efforts. However, this weekend I'm translating a text about milking and it insists on translating one keyword as "nipple", when the correct form is "teat" in this context. It's keeping me entertained...icon_smile.gif

You can actually use GT4T and other similar systems such as DeepL to translate entire pages or screeds of text, but then you would need to go through it and look for the inconsistencies or errors, so I don't normally use it that way.


 

Michael J.H. Davies  Identity Verified
Denmark
Local time: 02:15
Member (2009)
English to Danish
+ ...
Use of MT + post-editing Mar 10

I generally prefer MT as a first stage of translation using 2 different MT databases (directly in my CAT) to which I have subscriptions and which offer confidentiality of the text. MT saves a huge amount of typing and while the target text is far from immediately useable, I am convinced that it also saves time.

I use 2 stages of post editing - the first to make sure that the target text correctly reflects the content of the source text (but without worrying too much about grammar and spelling) and the second to correct grammatical and spelling errors and ensure that the text reads correctly as 'human translation' of the required quality.


 

Kevin Fulton  Identity Verified
United States
Local time: 20:15
German to English
Terminology and potential time savings Mar 10

Although I've not found Google Translate to be particularly useful for running text, single-term terminology searches often produce useful results due to the reasons offered by other posters on this topic.

Recently while translating a medical text, I was confronted with various lists of common symptoms (headache, nausea, dizziness, etc.) and anatomical terms most of which were easily recognizable but weren't in my terminology database. Copying and pasting saved me a lot of effort. This has proven to be true in other situations as well.

Translating whole sentences using MT hasn't been as useful, however, and the time it takes to fix the syntax makes it not a worthwhile endeavor, in my experience.


 

Gerard de Noord  Identity Verified
France
Local time: 02:15
Member (2003)
German to Dutch
+ ...
Let's not share Mar 11

For as long as I've been a member of ProZ.com I've wondered why people give away their trade secrets and bottom rates on this website. Let's for once ride the wave of being early adopters of innovative technology instead of being early adaptors to people who've found out too.

Cheers,
Gerard


 

Hans Lenting  Identity Verified
Netherlands
Member (2006)
German to Dutch
Nice example Mar 13

Hans Lenting wrote:

I cannot confirm that MT suggestions are always spelled correctly. Even good systems like DeepL create bogus words (by glueing words together). It would be nice to have an option to prevent word creation (word invention).


The seals are back:

wpg2usctzbupzggsmfke.png


 

Mandy Lim
United States
Local time: 17:15
Chinese to English
MT subscription Mar 13

May I ask what kind of MT database subscription that offers confidentiality you are using? Thank you.

Michael J.H. Davies wrote:

I generally prefer MT as a first stage of translation using 2 different MT databases (directly in my CAT) to which I have subscriptions and which offer confidentiality of the text. MT saves a huge amount of typing and while the target text is far from immediately useable, I am convinced that it also saves time.

I use 2 stages of post editing - the first to make sure that the target text correctly reflects the content of the source text (but without worrying too much about grammar and spelling) and the second to correct grammatical and spelling errors and ensure that the text reads correctly as 'human translation' of the required quality.


 

Robert Rietvelt  Identity Verified
Local time: 02:15
Member (2006)
Spanish to Dutch
+ ...
Zeehondenwater Mar 14

Hans Lenting wrote:

Hans Lenting wrote:

I cannot confirm that MT suggestions are always spelled correctly. Even good systems like DeepL create bogus words (by glueing words together). It would be nice to have an option to prevent word creation (word invention).


The seals are back:

wpg2usctzbupzggsmfke.png



I like that word, nice for a old Dutch word game, called 'Galgje'.icon_smile.gif


 

Robert Rietvelt  Identity Verified
Local time: 02:15
Member (2006)
Spanish to Dutch
+ ...
adopting or committing suicide? Mar 14

Gerard de Noord wrote:

For as long as I've been a member of ProZ.com I've wondered why people give away their trade secrets and bottom rates on this website. Let's for once ride the wave of being early adopters of innovative technology instead of being early adaptors to people who've found out too.

Cheers,
Gerard


Hi Gerard,

For once I agree with you, at least with your thesis 'Let's for once ride the wave of being early adopters of innovative technology'. Only I fear in this case that this 'innovative technology' will develop faster then we want it to and in the end will leave us without a job.




[Edited at 2018-03-14 20:50 GMT]

[Edited at 2018-03-14 20:51 GMT]


 


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


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

Sharing opinion about machine translation

Advanced search






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 »
SDL MultiTerm 2019
Guarantee a unified, consistent and high-quality translation with terminology software by the industry leaders.

SDL MultiTerm 2019 allows translators to create one central location to store and manage multilingual terminology, and with SDL MultiTerm Extract 2019 you can automatically create term lists from your existing documentation to save time.

More info »



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