I am using the latest version of 2019 and I have been experiencing this issue for a few weeks now. My main medical TM is no longer updating. It is about 1.3 GB. I am working with packages sent by my client, to which I append my own local TM. There is no other TM besides mine.
My issues are:
1) Newly translated segments are not being saved in the TM.
2) Segments are not leveraged correctly.
As for 1, I have checked the Update field is selected in the Project Settings. I have also tried to update the TM from Batch Tasks > Update TM (both Main TM and Project TM)
I get 0 errors but a tons of warnings due to 'an unspecified error' for each segment of my file.
I have also tried to use a language-specific TM. I have proof that the segments aren't saved in the TM because I open the TM in the TM Editor directly and it doesn't find the latest segments I have been translating.
I have tried to repair the TM using the TM repair tool and it was correctly 'repaired' and there is no error message whatsoever indicating something might be wrong.
As for 2, I don't understand why identical sentences are not recognized as such, like in the example below.
I am spending more time than due to translate because I have been working without TM. Basically, I filter segments and copy-paste the previous translation earlier in the same document. This is prehistoric!
Can you turn off lookahead and try again? Certainly the second problem seems to be related to this. I can't comment on TUs not being saved at all yet, but please try disabling lookahead first and then see what happen. It's under File->Options->Editor->Automation.
I always doubted the purpose of Lookahead, it has no advantages at all... I would simply remove this completely from future releases or at least make it DEACTIVATED by default. After a fresh Studio installation this is the first option I switch off...
Jerzy Czopik said:I always doubted the purpose of Lookahead, it has no advantages at all... I would simply remove this completely from future releases or at least make it DEACTIVATED by default.
I think I'd agree with you on the default. The usecase, at least the usecase with the least problems to look out for, is when you work only with MT and no terminology or TM at all. In this case the advantage of not having to wait for a response is tremendous. But until we have instantly adaptiveMT and people don't wish to use their other resources anymore the usecase for that is probably small to non-existent!
It can be quite useful, but we can see the number of problems people report that are based on not being able to retrieve results that have just been added, so it should certainly be an optional settings rather than the default... in my (and your) opion.
OK, this might be I do not have such a wide perspective as you have. In my usecases, which seem to be very similar to many other people, there is ALWAYS a TM and mostly a TB (very often even several TMs and TBs) in the project configuration. And even in case I would have the best possible MT on earth, I would still keep at least one TM and some TBs in the project settings.
To make this option useful, it must be configured that way, that in similar cases as here, when the subsequent segments to translate are similar and the next following segment would directly profit from the translation entered in just confirmed segment, the system recognizes this and makes a refresh in the search despite the prefetched results. If lookahead would already have worked that way, there wouldn't have been any compliants about it.
Jerzy Czopik said:And even in case I would have the best possible MT on earth, I would still keep at least one TM and some TBs in the project settings.
As I said... "But until we have instantly adaptiveMT and people don't wish to use their other resources anymore the usecase for that is probably small to non-existent!"
We're in complete agreement.