Yevgnen/pybrat

discontinuous text-bound annotations not working

zeljkobekcic opened this issue · 4 comments

zeljkobekcic commented

Hello,

I ran into a problem with the regex which parses entities (https://github.com/Yevgnen/pybrat/blob/main/pybrat/parser.py#L110).
The other regexes are affected by this issue too.

Brat (from v1.3) allows a semicolon if the annotation continues in the next line.
The official documentation provides an example for this https://brat.nlplab.org/standoff.html.

zeljkobekcic commented

Changing the regex to something like:

(?P<id>T\d+)\t(?P<type>[^ ]+)\ (?P<start>\d+)\ (?P<end>\d+)(;\d+\ \d+)*?\t(?P<mention>.+)

Here I added an optional group which captures the other two indices.

will accept the test string T1 Location 0 5;16 23 North America from the example.

Yevgnen commented

Hi, thanks for reporting the issue and the fix. Can you check if this fix work for you? Note that since there may be multiple spans for an entity, the .start and .end field will now only work continuous entities.

zeljkobekcic commented

I will report you later if it fixes it. Thank you very much for you quick reply!

zeljkobekcic commented

It took a little longer to get back to it, but it looks like it's working!
Thank you very much!

  • 🎉1