Thank you Philippe, this is a really great showcase of ModernMT capability, and it proves a deep understanding of the inners of our tool. Thank you for the amazing work and we literally cannot wait to see ModernMT perfectly integrated in Wordfast!
URL: api.modernmt.com/translate?source=en&target=fr&q={ss} (replace with your language code, here en and fr) JSON key: translation Header: MMT-ApiKey=8888888888888,X-HTTP-Method-Override=GET (replace all the "8" with your API key)
Great video and well put together. In our practice, we use MemoQ. Originally, we chose DeepL as an MT service provider. A year ago, we switched to ModernMT. After over a million words translated, we are, on the whole, mostly satisfied with ModernMT. We are pleased with the adaptive features, especially with verb tenses. However, we're less than happy with terminology results. One of our uploaded medical TM contains more than 50K segments and is very clean. For example, the source term "dorsolombaire" and all its variants are systematically translated as "thoracolumbar" in that particular TM. After using the pretranslate functionality in MemoQ, ModernMT returns as a translation "dorsolumbar" "dorso-lumbar" dorsolombar" or "spinal" !! How disappointing! I have so many examples like this one. We end up having to run a large number of regular expressions to find and replace those unfortunate term translations. I'm about to write a long email to ModernMT about the issue. Have you experienced something similar? On a different topic, if you are familiar with regular expressions, it might be a good topic for numerous videos. There now seems to be a trend for translators to get familiar with REGEX.
Thank you so much for this comment, very useful! For such situations, what I would do is create a training file with 10 different sentences containing the term you wish to be translated differently. Translate and fix the issues, like I show in the later part of the video, this should be recognized by ModernMT and translate this specific word according to your wishes when this cloud TM is active in MMT. Custom MT is a bit of a pain. However, some MT providers have options to force glossary on MT output. If the above solution doesn't fix you problem, I know Intento have a dedicated system for that in their mid-tier price. You could bring your MMT key to their platform, and then just pay for the service + glossaries. You terminology preferences would then be applied on MT output. Regarding REGEX, yet this topic is becoming frequent, but depending on the software you are feeding REGEX into, the way of writing it changes because they rely on different REGEX "flavors". My experience is that for translators at large, it's easier to have a bank of ready-made, copy-paste expressions to get specific results, maybe someone would like to publish a library of these on the web sorted by CAT and by desired result, that would be amazing!
Thanks Philippe for creating this video! I too use ModernMMT - not in SDL but in memoQ where, sadly, the context analysis at document level doesn't seem to be available (yet).
Hello, Thank you for your comment Patrick. You are absolutely correct! At the moment this context analysis feature is working fully only in SDL and Matecat. I hope it will be added to other CAT tools soon like MemoQ and Wordfast Pro. Hopefully MemoQ will find this video and fix this feature for you. My bests
Hi Patrick! Unfortunately, this is correct and it comes from a limitation of the memoQ integration APIs. However, if you (and anyone interested) could send an email to memoQ asking for this feature, they will eventually update the API and let us integrate the context analysis.
So did you see difference between loop version vs professional one . İt is said enterprise version learns from corrections of translator,is it really so? Which costs 50 usd
Yes that's what I demonstrate in the video at the end forcing the MT to suggest a term. The way they do it is via a cloud TM that you can upload and you can feed directly as you translate. MT output is compared against TM when there are matches before it is sent to the translator. So there really is an adaptation to the translator. This will work best is you always translate the same things. If not, I suggest you create different TMs in the ModernMT plugin and use each time the one that fits what you will be translating. Hope this helps.
Hi M Celique! ModernMT "Human-in-the-loop" and "Professional" plans differ only on how many users can use the plan and the billing. "Professional" is intended for professional translators, so it has unlimited personal use with a 25$ fee per month. "Human-in-the-loop" is more intended for LSPs so there is no limitation on how many users can use it and the billing is totally a pay-per-use model: 50$ is per million chars, if you translate 100,000 chars, you get billed 5$, if you translate 2,000,000 chars, you the bill will be 100$ and so on. Most importantly, in terms of features and adaptation, Human-in-the-loop and Professional are exactly the same: they both adapt in real-time via TMs and even single post-edits, full CAT Tool support, and context analysis. If you have any questions regarding the product, please feel free to contact ModernMT at support@modernmt.com!
Hello! I have been using Modern MT in almost a year and I am quite satisfied with the results in comparison to other machine translation tools. Unfortunately, it is not always effective. I have noticed that it sometimes needs the same sentence translated several times in the same way to get what is the correct translation. I can not understand why it wont do it after only 1 time (?). The other thing that is disappointing is when I use it with a new project it sometimes goes one step back and again translates incorrect !! thoug it has got enough feedback in the past on how the particular word or a sentence is supposed to be translated in a very similar context. I have also been missing glossary BUT it turnes out that this fall ModernMT finnally got a glossary feature BUT I find it too complicated that I have not been able to use it and I dont find any videotutorial on this. The technical description on their page is extremely confusing :( It also seems to have some unusual format which I dont understand. Whould you be so kind and check this new feature out? and make a video on it? I would appreciate it a lot :-) Greetings from a fellow translator from Norway!
Hi, yes it can feel strange. In the video I had to force a word 10 times to get it to use it. Can you put the link for the page referring to the glossary in your next message? (write "DOT" instead of pasting the url with the dots otherwise you comment will be discarded by RUclips).
Thank you Philippe, this is a really great showcase of ModernMT capability, and it proves a deep understanding of the inners of our tool. Thank you for the amazing work and we literally cannot wait to see ModernMT perfectly integrated in Wordfast!
Thanks for your kind comment. I hope this will help users understand better what they can get out of the excellent tool your team built!
Hello to all. It is now possible to connect ModernMT directly in WF Pro via the CustomMT fields, fill up as described in next comment:
URL: api.modernmt.com/translate?source=en&target=fr&q={ss} (replace with your language code, here en and fr)
JSON key: translation
Header: MMT-ApiKey=8888888888888,X-HTTP-Method-Override=GET (replace all the "8" with your API key)
Great video and well put together. In our practice, we use MemoQ. Originally, we chose DeepL as an MT service provider. A year ago, we switched to ModernMT. After over a million words translated, we are, on the whole, mostly satisfied with ModernMT. We are pleased with the adaptive features, especially with verb tenses. However, we're less than happy with terminology results. One of our uploaded medical TM contains more than 50K segments and is very clean. For example, the source term "dorsolombaire" and all its variants are systematically translated as "thoracolumbar" in that particular TM. After using the pretranslate functionality in MemoQ, ModernMT returns as a translation "dorsolumbar" "dorso-lumbar" dorsolombar" or "spinal" !! How disappointing! I have so many examples like this one. We end up having to run a large number of regular expressions to find and replace those unfortunate term translations. I'm about to write a long email to ModernMT about the issue. Have you experienced something similar? On a different topic, if you are familiar with regular expressions, it might be a good topic for numerous videos. There now seems to be a trend for translators to get familiar with REGEX.
Thank you so much for this comment, very useful! For such situations, what I would do is create a training file with 10 different sentences containing the term you wish to be translated differently. Translate and fix the issues, like I show in the later part of the video, this should be recognized by ModernMT and translate this specific word according to your wishes when this cloud TM is active in MMT. Custom MT is a bit of a pain. However, some MT providers have options to force glossary on MT output. If the above solution doesn't fix you problem, I know Intento have a dedicated system for that in their mid-tier price. You could bring your MMT key to their platform, and then just pay for the service + glossaries. You terminology preferences would then be applied on MT output. Regarding REGEX, yet this topic is becoming frequent, but depending on the software you are feeding REGEX into, the way of writing it changes because they rely on different REGEX "flavors". My experience is that for translators at large, it's easier to have a bank of ready-made, copy-paste expressions to get specific results, maybe someone would like to publish a library of these on the web sorted by CAT and by desired result, that would be amazing!
Thanks Philippe for creating this video! I too use ModernMMT - not in SDL but in memoQ where, sadly, the context analysis at document level doesn't seem to be available (yet).
Hello, Thank you for your comment Patrick. You are absolutely correct! At the moment this context analysis feature is working fully only in SDL and Matecat. I hope it will be added to other CAT tools soon like MemoQ and Wordfast Pro. Hopefully MemoQ will find this video and fix this feature for you. My bests
Hi Patrick! Unfortunately, this is correct and it comes from a limitation of the memoQ integration APIs. However, if you (and anyone interested) could send an email to memoQ asking for this feature, they will eventually update the API and let us integrate the context analysis.
So did you see difference between loop version vs professional one . İt is said enterprise version learns from corrections of translator,is it really so? Which costs 50 usd
Yes that's what I demonstrate in the video at the end forcing the MT to suggest a term.
The way they do it is via a cloud TM that you can upload and you can feed directly as you translate. MT output is compared against TM when there are matches before it is sent to the translator.
So there really is an adaptation to the translator.
This will work best is you always translate the same things. If not, I suggest you create different TMs in the ModernMT plugin and use each time the one that fits what you will be translating. Hope this helps.
Hi M Celique!
ModernMT "Human-in-the-loop" and "Professional" plans differ only on how many users can use the plan and the billing. "Professional" is intended for professional translators, so it has unlimited personal use with a 25$ fee per month. "Human-in-the-loop" is more intended for LSPs so there is no limitation on how many users can use it and the billing is totally a pay-per-use model: 50$ is per million chars, if you translate 100,000 chars, you get billed 5$, if you translate 2,000,000 chars, you the bill will be 100$ and so on.
Most importantly, in terms of features and adaptation, Human-in-the-loop and Professional are exactly the same: they both adapt in real-time via TMs and even single post-edits, full CAT Tool support, and context analysis.
If you have any questions regarding the product, please feel free to contact ModernMT at support@modernmt.com!
Thanks Davide, so cool to get that kind of detailed answer!!!
Hello! I have been using Modern MT in almost a year and I am quite satisfied with the results in comparison to other machine translation tools. Unfortunately, it is not always effective. I have noticed that it sometimes needs the same sentence translated several times in the same way to get what is the correct translation. I can not understand why it wont do it after only 1 time (?). The other thing that is disappointing is when I use it with a new project it sometimes goes one step back and again translates incorrect !! thoug it has got enough feedback in the past on how the particular word or a sentence is supposed to be translated in a very similar context. I have also been missing glossary BUT it turnes out that this fall ModernMT finnally got a glossary feature BUT I find it too complicated that I have not been able to use it and I dont find any videotutorial on this. The technical description on their page is extremely confusing :( It also seems to have some unusual format which I dont understand. Whould you be so kind and check this new feature out? and make a video on it? I would appreciate it a lot :-) Greetings from a fellow translator from Norway!
Hi, yes it can feel strange. In the video I had to force a word 10 times to get it to use it. Can you put the link for the page referring to the glossary in your next message? (write "DOT" instead of pasting the url with the dots otherwise you comment will be discarded by RUclips).