My model misses span boundaries and I'm fixing it by reviewing annotations as part of dataset bootstrap, but its really annoying to first delete entity annotation and then dragging over the correct span again
Is it possible to have "extend" feature in UI when I can "add" tokens to the labeled span somehow?
Few clicks saved here and there would make a huge difference - eg i would "drag" from the "acquisitions" to the right until I mark the whole constituent and accept the annotation 
hi @ysz!
Thanks for the feedback. I really like how you're thinking about making annotation more efficient.
Can you explain more about what you mean by an "extend" feature?
By "add tokens to the labeled span" -- are you suggesting having some sort of input text box where you can type in the full span of what you want? I'm not following how you're envisioning to "add tokens" instead of clicking/dragging.
Thanks for getting back to me @ryanwesslen , so the model pre-annotates Spans, right. Span can be "the transformative" out of ...the transformative acquisition... where clearly model missed to include NP into a larger and more meaningful NP "chunk"
In prodigy it will have a box around the model prediction ie
... [the transformative] acquisitions...
SOME_LABEL
What if I can just drag the right border of this "box" in order to start "including" tokens to the right into the annotation so
... [the transformative >>>dragging this boundary to the right>>] acquisitions...
SOME_LABEL
then
... [the transformative acqui >>>keep dragging and boundary crosses next token boundaries>>] sitions...
SOME_LABEL
finally
... [the transformative acquisitions >>>releasing mouse somewhere here or near the border of last token>>> ]...
SOME_LABEL
makes it into
... [the transformative acquisitions ]...
SOME_LABEL
where SOME_LABEL is now applied to the chunk of 3 tokens
Let me know if it explains idea better!
hi @ysz!
Sorry for the delay, but thanks for the description. That was very helpful. I've created an internal ticket so we can track this request.
The team is busy with v1.12 patches and working hard on v1.13 too, so we'll need to prioritize it accordingly. However, I'll post back if we can add this on a future release.
Thanks again for your feedback!