Adaptive Machine Translation
Thread poster: Manuela Ticudean
Manuela Ticudean
Manuela Ticudean
Romania
English to Romanian
+ ...
May 7, 2019

Hi everyone!

I'm a bachelor student in translation, it is my final semester and I'm working on my final thesis. I have a chapter dedicated to machine translation and I'd like to write about adaptive MT as well, but I wasn't able to find too many resources about it.

I'm interested in whether it "improves" the experience of working with machine translation in any significant way (from a translator's point of view). Unfortunately I don't have any real personal experience
... See more
Hi everyone!

I'm a bachelor student in translation, it is my final semester and I'm working on my final thesis. I have a chapter dedicated to machine translation and I'd like to write about adaptive MT as well, but I wasn't able to find too many resources about it.

I'm interested in whether it "improves" the experience of working with machine translation in any significant way (from a translator's point of view). Unfortunately I don't have any real personal experience with post-editing or machine translation.

I noticed that there was a bit of a hype about it in 2016 and 2017, but then 2018 has been really quiet about this topic. The EUATC language survey of 2018 and 2019 don't mention adaptive MT in any way (or maybe it is just included in the general MT term).

I would appreciate any opinions, thoughts or insights into how it is to work with this technology.

Thanks a lot!
Manuela
Collapse


Philippe Locquet
 
Manuela Ticudean
Manuela Ticudean
Romania
English to Romanian
+ ...
TOPIC STARTER
Recent video from Lilt May 8, 2019

While continuing my research today I stumbled on this video: Recent Progress in Interactive Neural Machine Translation by Lilt.
It is precisely on interactive MT and it's from early 2019. Maybe somebody else will find it interesting.

My conclusion from watching it is that Adaptive MT is still very new and probably not many translators have had access to it.

I would still appreciate any comme
... See more
While continuing my research today I stumbled on this video: Recent Progress in Interactive Neural Machine Translation by Lilt.
It is precisely on interactive MT and it's from early 2019. Maybe somebody else will find it interesting.

My conclusion from watching it is that Adaptive MT is still very new and probably not many translators have had access to it.

I would still appreciate any comments or opinions, thank you!


[Edited at 2019-05-08 18:08 GMT]
Collapse


 
Philippe Locquet
Philippe Locquet  Identity Verified
Portugal
Local time: 12:34
English to French
+ ...
Some info out there May 8, 2019

Manuela Ticudean wrote:

Hi everyone!

I'm a bachelor student in translation, it is my final semester and I'm working on my final thesis. I have a chapter dedicated to machine translation and I'd like to write about adaptive MT as well, but I wasn't able to find too many resources about it.

I'm interested in whether it "improves" the experience of working with machine translation in any significant way (from a translator's point of view). Unfortunately I don't have any real personal experience with post-editing or machine translation.

I noticed that there was a bit of a hype about it in 2016 and 2017, but then 2018 has been really quiet about this topic. The EUATC language survey of 2018 and 2019 don't mention adaptive MT in any way (or maybe it is just included in the general MT term).

I would appreciate any opinions, thoughts or insights into how it is to work with this technology.

Thanks a lot!
Manuela


Hi Manuela,

This is an interesting project and the way this operates is still taking shape in the industry.

There are a couple of ways this is approached (non-extensive list):
1_LSP uses Custom engines for a number of clients.
2_LSP uses a special engine for a special client.
3_Company subscribes to the service for in-house translations and get a Custom engine that helps in-house translators (fast translation for low-cost is the goal).
4_Same as 3 but pretranslated from the engine is sent to translators for "proofreading" (knowingly or not that it was AI translated). No comment.

Then translators can be involved in engine improvement, this is a specific task where translators are asked for special editing. This serves then to retrain the engine so that it performs better. PEMT or MTPE.

There is also a root approach ensuring what is fed to train engines is top-quality translations, this reduces future MTPE efforts.

Translators have been involved in this always, since engines piggyback on what has been previously translated by humans to be trained accordingly trying to replicate human translation.

There have been big discussions on such topics on ProZ.com here are two threads I can remember:
https://www.proz.com/forum/business_issues/318053-machine_translation_–_post_editing_should_we_take_on_these_jobs_for_how_much.html
https://www.proz.com/forum/post_editing_machine_translation/330203-custom_machine_translation_a_new_revolution_on_the_way.html

I have published a few videos on this subject, here are the most current:
https://youtu.be/0YeyhA4Jnv4
https://youtu.be/0Ta2dXRTvU4

Hope this gives you a few leads for your work.

My bests


 
Manuela Ticudean
Manuela Ticudean
Romania
English to Romanian
+ ...
TOPIC STARTER
Very useful May 9, 2019

This is extremely useful information for me and I have to take some time to carefully go through everything.

I was aware of the first forum discussion you mentioned, but not the second.

Thank you for taking the time and linking the videos.

Manuela


 
Tom Hoar (X)
Tom Hoar (X)
United States
Local time: 08:34
English
Study of Machine Translated Segment Pairs Aug 16, 2019

Hi Manuela,

The "adaptive" approach to machine translation seems to have gone out of style. Less than a year after companies promoted it, neural machine translation (NMT) became the newest "best thing since sliced bread."

About that time, I did a commercial research project that compared Google's new NMT to customized statistical machine translation (SMT). The results show that for professional translators, the actual NMT technology didn't live up to the marketing hype
... See more
Hi Manuela,

The "adaptive" approach to machine translation seems to have gone out of style. Less than a year after companies promoted it, neural machine translation (NMT) became the newest "best thing since sliced bread."

About that time, I did a commercial research project that compared Google's new NMT to customized statistical machine translation (SMT). The results show that for professional translators, the actual NMT technology didn't live up to the marketing hype that accompanied its introduction to the market.

Study of Machine Translated Segment Pairs

Maybe this will give you some ideas for your thesis.

Tom
Collapse


 


There is no moderator assigned specifically to this forum.
To report site rules violations or get help, please contact site staff »


Adaptive Machine Translation







Anycount & Translation Office 3000
Translation Office 3000

Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.

More info »
Trados Business Manager Lite
Create customer quotes and invoices from within Trados Studio

Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.

More info »