using ModernMT with memoQ Thread poster: Angel Llacuna
|
Do you use ModernMT with memoQ? Does the adaptive learning of that tool work well? I would like to receive some feedback from ModernMT users before buying the product. link : https://www.modernmt.com/ | | | Post removed: This post was hidden by a moderator or staff member for the following reason: User request. |
I have tried it with different text types and it does work and it does learn. It can be easily configured from memoQ with translation memories specific for your clients/project types and it works reliably. I have tried EN>ES and DE>ES, with smaller tests with other combinations, like FR>ES. The quality (also in idiomaticity, language variety, depth of comprehension, collocations...) I have experienced is superior to that of Deepl, especially when it comes to a trained configuration ... See more I have tried it with different text types and it does work and it does learn. It can be easily configured from memoQ with translation memories specific for your clients/project types and it works reliably. I have tried EN>ES and DE>ES, with smaller tests with other combinations, like FR>ES. The quality (also in idiomaticity, language variety, depth of comprehension, collocations...) I have experienced is superior to that of Deepl, especially when it comes to a trained configuration (which Deepl cannot offer at present). Four important caveats though: 1. When you load a memory (in the form of a TMX), try not to use big memories or, at least, if you use a big memory, make sure it is specific for the project at hand. ModernMT needs to find your preferred terminology easily, and variability in terminology within the TM makes it harder for ModernMT to hit the right note. 2. ModernMT does not learn on the first go, i.e. it can happen that you have to correct the same term several times until ModernMT prioritizes it vs. the standard terminology in its turnkey, untrained corpus, and it can also happen that, once ModernMT has learnt your preference, it reverts to the standard terminology occasionally. 3. For non-English languages, please notice (I asked the ModernMT team about a year ago, so this might have changed) that English is used as an interlingua. In my case, if I translate German into Spanish, German is internally translated into English and only that into Spanish. The process is very much instantaneous, however, and from what I have seen there is not much of a loss in terms of idiomaticity or pragmatics. 4. (This is general in memoQ with any MT systems): Please be aware that memoQ can mark your use of MT in XLIFF files you deliver. It happens, for instance, in SDL Studio files (SDLXLIFF files) translated with memoQ. If this is a concern for you for whatever the reason, then MT in memoQ is a no-go. (I can tell you more about it if you email me directly; if would be a bit long to explain here.) All in all, I would favour ModernMT to Deepl, and the cost per month is very similar indeed, I would say. ▲ Collapse | | | One more thing: Productivity | May 1, 2021 |
Just one thing I would like to mention, if only to set sensible expectations: even if it is quite good, ModernMT will not speed you up dramatically overnight. I would say that, for complex texts, your work as a translator is as required as usual, and ModernMT will serve as a way to spare a bit of typing. You still need to revise everything and do your translation mentally, so it might save you some effort, but it does not replace you by any means. In my tests and measurements with M... See more Just one thing I would like to mention, if only to set sensible expectations: even if it is quite good, ModernMT will not speed you up dramatically overnight. I would say that, for complex texts, your work as a translator is as required as usual, and ModernMT will serve as a way to spare a bit of typing. You still need to revise everything and do your translation mentally, so it might save you some effort, but it does not replace you by any means. In my tests and measurements with ModernMT, a trick that worked for me is that I read the source text and thought of how I would have translated the sentence manually. Only then did I look at what ModernMT had produced and changed its proposal to my taste and preference. ▲ Collapse | |
|
|
Angel Llacuna Spain Local time: 09:20 English to Spanish TOPIC STARTER thank you, Tomás | May 2, 2021 |
In my case, I would use ModernMT for translations from English to Spanish, with texts of Information Technology and I must adhere to the Microsoft glossary. Does ModernMT "learn" terminology from the translations that I wrote or it is necessary to load a translation memory first ?
[Edited at 2021-05-02 05:57 GMT] | | | An alternative could be a free offline NMT-Engine | May 3, 2021 |
In the Issue 21-4-324 of A Computer Journal For Translation Professionals (internationalwritersdotcomslashtoolkitslashcurrentdothtml) you can find a detailed description, links and the author’s opinion on the free offline Neural Machine Translation (NMT) Engine OPUS-CAT. Plugins for Studio, memoQ, Wordfast and OmegaT are available. I can only confirm that the engine works fine with the plugin for Studio, but had not time to fine-tune (train) the engine or to install the plugin i... See more In the Issue 21-4-324 of A Computer Journal For Translation Professionals (internationalwritersdotcomslashtoolkitslashcurrentdothtml) you can find a detailed description, links and the author’s opinion on the free offline Neural Machine Translation (NMT) Engine OPUS-CAT. Plugins for Studio, memoQ, Wordfast and OmegaT are available. I can only confirm that the engine works fine with the plugin for Studio, but had not time to fine-tune (train) the engine or to install the plugin in memoQ. In addition, the engine allows a pre-translation. As far as the use of the MS accepted terminology concerns, you can attach the MS glossary to the TBs used in memoQ and you will always have a second control (beside the TM used for training) on the terminology used. IMO the use of an own offline MT engine, trained based on your own TMs (containing the MS terminology) cannot be interpreted as a breach of the NDA. You can give it a try. ▲ Collapse | | | Angel Llacuna Spain Local time: 09:20 English to Spanish TOPIC STARTER Thanks George .. | May 3, 2021 |
for your detailed answer. I will take a look at the mentioned document | | | Would need to test it | May 3, 2021 |
George Luca wrote: In the Issue 21-4-324 of A Computer Journal For Translation Professionals (internationalwritersdotcomslashtoolkitslashcurrentdothtml) you can find a detailed description, links and the author’s opinion on the free offline Neural Machine Translation (NMT) Engine OPUS-CAT. This is an interesting suggestion, but I think it would require some testing. I have tried building my own MT engines just for the sake of it, and the results were rather poor. I think that the combination of a powerful generic MT + training with your own resources and on-the-fly learning capabilities is the way to go and will be the future. The reason why I do not think an in-house engine is the solution is because it would lack the corpora content or the power to produce idiomatic translations with sensible collocations and solid handling of ambiguous sentences, one typical example being the ubiquitous English be+participle sentences that can be understood as a state or as a change of state ("The device is disabled" -> "The device is in a disabled state" vs. "The device gets disabled at this moment"). Such ambiguous sentences occur in different languages, and only a neural system with varied, curated corpora is more prepared to handle these situations well. In any case, it would be interesting to put OPUS-CAT to a test! | |
|
|
Michael Beijer United Kingdom Local time: 08:20 Member (2009) Dutch to English + ... ‘Cattitude’ (brand new LOCAL adaptive machine translation system about to be released) | May 3, 2021 |
Just thought I'd mention another (very interesting) adaptive machine translation system that is about to be launched. A bit of background: ModernMT is open source (it lives here: https://github.com/modernmt/modernmt ), meaning … you can take the source code and tweak it. This is exactly what Loek van Kooten (an English/Japanese-Dutch game translator, and owner of Loekalization.com) did, and ... See more Just thought I'd mention another (very interesting) adaptive machine translation system that is about to be launched. A bit of background: ModernMT is open source (it lives here: https://github.com/modernmt/modernmt ), meaning … you can take the source code and tweak it. This is exactly what Loek van Kooten (an English/Japanese-Dutch game translator, and owner of Loekalization.com) did, and he is working on releasing his very interesting project under the name ‘Cattitude’. Have a look: • https://c4ttitude.com/ • https://www.loekalization.com/modernmt.html • https://www.youtube.com/watch?v=sZFawemcn74 (make sure to watch this!) ▲ Collapse | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » using ModernMT with memoQ Wordfast Pro | Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
| TM-Town | Manage your TMs and Terms ... and boost your translation business
Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |