You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have some questions about "good management" practice with regex rules and how to best manage them.
Here are some observations - hence the question marks - for you to comment on please to say if I'm right or wrong, is there a better method or that I have misunderstood something.
I should NOT create them with a trained MT engine as I have to retrain such engines after every job and I risk losing them when I delete the old trained engine?
So I should perhaps save them to a with an installed model that I then use for training but would they be automatically transferred to the new trained MT engine?
I started to play around with all this and exported a few rules and collections to see if I could reimport them into another trained MT engine but here I got a real shock as when I looked at the collections/rules that I had saved to my special folder, they had names like "63e8e851-3936-49f8-9c3d-125476b5033e.yml" . And NotePad ++ does not seem to like opening them. It takes a long time but once open, I can see what each file contains. But on reimport, I have to remember or note down what all these files contain and then hunt for the right one(s). This was mind blowing and I only had a few collections/rules on my desktop
Furthermore, there was no clear sign in the file as to what installed model the rule/collection had come from (French or Swedish to English). So I guess that I should always put this in the name of the regex rule/collection name and perhaps save them to different sub-folders in future
To resume, this was just a first test run and I hope to get better, but I can see that creating rules and collections might need careful planning and I'm particularly concerned about the long "non-transparent" names individual rules and collections are given, making reimport very confusing
Any advice please?
Thanks in advance
Dave Neve
The text was updated successfully, but these errors were encountered:
I think I can get around the other problems too if we can add to or create the names. Then we can add stuff like FR (French) and SE (Swedish) as some rules have exactly the same description as they do the same thing, but the regexes themselves are different due to the layout of the languages (such as with numbers)
Do you think this enhancement will be ready by next Monday ? (English humour 😂😂😂)
Hello Tommi and all
I have some questions about "good management" practice with regex rules and how to best manage them.
Here are some observations - hence the question marks - for you to comment on please to say if I'm right or wrong, is there a better method or that I have misunderstood something.
I should NOT create them with a trained MT engine as I have to retrain such engines after every job and I risk losing them when I delete the old trained engine?
So I should perhaps save them to a with an installed model that I then use for training but would they be automatically transferred to the new trained MT engine?
I started to play around with all this and exported a few rules and collections to see if I could reimport them into another trained MT engine but here I got a real shock as when I looked at the collections/rules that I had saved to my special folder, they had names like "63e8e851-3936-49f8-9c3d-125476b5033e.yml" . And NotePad ++ does not seem to like opening them. It takes a long time but once open, I can see what each file contains. But on reimport, I have to remember or note down what all these files contain and then hunt for the right one(s). This was mind blowing and I only had a few collections/rules on my desktop
Furthermore, there was no clear sign in the file as to what installed model the rule/collection had come from (French or Swedish to English). So I guess that I should always put this in the name of the regex rule/collection name and perhaps save them to different sub-folders in future
To resume, this was just a first test run and I hope to get better, but I can see that creating rules and collections might need careful planning and I'm particularly concerned about the long "non-transparent" names individual rules and collections are given, making reimport very confusing
Any advice please?
Thanks in advance
Dave Neve
The text was updated successfully, but these errors were encountered: