My first machine translation post-editing assignment
After an endless exchange of confirmations with the client – you know how it goes, they ask you to confirm and then re-confirm after you have confirmed every single detail of the project and in the end, you feel like maybe you should have confirmed once again, just for the sake of doing it – there it is, my first MTPE assignment.
Subject matter: IT (Software Documentation)
Volume: 250 words
Turnaround time (TAT): 1 day
This should be easy because my productivity is at its finest in the fields of IT and Legal. While my regular norm is 3,000 words per day, it can go up to 10,000 words in these two fields. I cannot tell for sure what my daily norm is in other fields, but I know that I can translate 1,000 words in an hour in any working field provided that:
- I am familiar with the tool which allows for spellchecking as you type and automated QA check (hello memoQ, I love you!)
- I am familiar with the subject matter and instructions are clear
- Terminology is readily available in the form of Translation Memory (TM), termbase or glossary and I don’t need to look it up in external online resources
- I can type at my usual rate of 55 words per minute without distractions (mind you, I do have a cat and a dog) and without technical glitches
- Source text is clearly written without superfluous tags
So it would take me anything between 15-30 minutes to complete regular translation of 250 words in the field of IT.
This being my first MTPE assignment, I am obviously not familiar with the tool. It is a cloud-based online translation tool where you have to sign in with credentials previously provided by the client. Neural machine translation engine is incorporated in the tool, so the suggested translations of source segments are already provided when you log in.
First things first, it’s an online tool so I am pretty sure there is no dictionary for Croatian and I can say goodbye to spellchecking as I type…
So I ask the client's Project Manager (PM) about the spellchecker and they tell me: the tool does not have an integrated spell checker, rather it uses the browser’s own. Yeah, right, not for Croatian. This instantly means that I will spend more time on this task because I will have to reduce my typing speed so as to avoid typos and I will have to re-read the translation twice to make sure no typos were overlooked. Aight, it’s 250 words, I can live with that.
The tool itself looks just like any other CAT tool today; at the top there is a box with the source text, another box for target segment (already populated with MT), where the translator (post-editor, PE?) edits the pre-translated text. Once you press Save, you move on to the next segment. That’s nice, I hate clicking around only to move to the next segment.
There is a window for Translation Memory matches, but the TM seems to be empty. There is also a Termbase window, but it also seems to be empty. Another good thing about the tool is that it propagates repetitions, but alas, there is no QA check! What if I inadvertently left a trailing space at the end of a segment or used incorrect quotation marks? I would never forgive myself. This asks for another re-read (adding up to 3?)
Well, what can I say; it came as no big surprise that content is rubbish. This thing is an IT-related text, filled with Microsoft terminology, but one can tell at a glance that none of the used Croatian terms were MS approved. In case you didn’t know, there is a free online Microsoft Terminology Collection available at https://www.microsoft.com/en-us/language. It is a set of standard technology terms used across Microsoft products. If you ask me, it used to be cooler before the Windows 8 edition because now it seems full of inconsistencies, but that's a whole 'nother story. When you have references to Microsoft products, you must use it in your translation.
So, I have an IT-related text filled with Microsoft terminology, but no termbase, no TM and obviously the machine used had not been introduced to Microsoft Terminology prior to its translation. OK, no biggie, I’m pretty familiar with MS terminology anyway, so that should not be a problem.
There is a whole lot of user interface (UI) and help content items to be translated, so I need to perform another check with the PM as to whether these need to be fully translated or handled bilingually (e.g. Press OK > Pritisnite OK (U redu) or Pritisnite U redu (OK)). I did not receive any instructions on how to handle these and this is of vital importance because some interfaces are localized and some are not and you as a translator need to know what the final product will look like...
Anyway, stay tuned, I will go further into content once I have all the info.