Assorted comments/questions
Closed this issue · 2 comments
How do you create a ta or tajson file from a text file? Is there a config format for plain text?
If I click on highlighted entity, e.g. Miss Watson in doc-EnglishTa.txt and select a different label 'ORG', it doesn't change color or appear in the log. It does appear to change for things I've annotated myself.
It might be nice to remove a label in the label selection dialog (easier to find/remember than right clicking if you have closely read the readme).
Is the idea that we can reconstruct the labels by replaying the log? I can find the orig-annotation-user files, but how do I read them?
Where do initial labels come from? They are not in the log, so I'm not clear how to get them.
What happens when you submit a definition?
I'll answer each in order:
Use txt2tajson.py. There is no config format for plain text because we always save to the format we load, and there's no established way of storing NER annotations in plain text.
Are you clicking on just one token? (e.g. Watson, or Miss). If you click and drag (start on Miss, hold down, move to Watson) then select label it should work. If this doesn't work, please open a separate issue with steps to reproduce.
There used to be a Non-entity button, but I removed it. Thanks for the suggestion, we'll consider for future updates.
The log is intended as a log, not necessarily for recreating. I'm not sure I understand the question about the orig-annotation-user files.
Input data may be labeled. In this case, I guess only the ta/eng data has labels (sorry, I can see how that is confusing), but any input data may contain labels.
This should be stored in a local dictionary, and also written to file. Once you add a definition, you should be able to check "Show dictionary" and all definitions will be shown inline. This appears to not work at the moment. Thanks for pointing it out.