My initial thought was to have a CasConsumer that loads annotations
directly into Knowtator programmatically, and a CasInitializer that goes
the other way. What remains is to have a way to translate/synchronize
the Type System in UIMA with the class hierarchy / annotation schema in
Knowtator back and forth. Both of these tasks should be fairly straight
forward. I would rather do it this way than to muck with translating
file formats.
Making it possible to go from one to the other is one thing but making
it fun and easy is another. Some considerations are:
- Knowtator has extra book keeping information that it associates with
each annotation such as the annotator, the creation date, and comments
- Knowtator allows multiple spans for a single annotation
- Knowtator allows you to annotate mentions of classes (e.g. person) or
mentions of instances (George Washington)
... and there are always usability issues, etc.
see other answers interspersed below.
Thilo Goetz wrote:
Hi Philip,
I downloaded Knowtator and played around with it a bit. It looks pretty
slick (although I think it could be improved by following some of the more
standard UI conventions, like dialogs with "ok" and "cancel" buttons, for
example; though I guess that's Protege, not Knowtator).
Yes. Protege uses non-modal dialogs by design. There is a lengthy
explanation
<http://protege.stanford.edu/doc/design/ok_and_cancel_buttons.html> of
this on the Protege website.
Also note that we have a contribution by Joern Kottmann in the sandbox called
"CAS editor". This is Eclipse based tooling that also allows you to manually
create annotations. It is still under development, but maybe there could
be some cross-fertilization.
The CAS editor is very hard to find! You might consider putting a link
on the sandbox page. I will take a look at this. If nothing else,
Knowtator might serve as a source of requirements / feature requests for
the CasEditor. Also, calculating IAA is non-trivial for a variety of
reasons which deserve their own conversation. Do you have any notion
what Joern's commitment is to the project?
[I know it's a bit early to talk about legal issues, but please note that
the Mozilla license is incompatible with the Apache license. If any of your
code were to move to Apache, it would need to be relicensed under ASL 2.0.]
We chose MPL because that is what Protege uses.
Re funding: we are optimistic that there will be more UIMA innovation
awards by IBM this year. Watch this space for the announcement. I would
think that a Knowtator/UIMA integration could be a candidate.
Great! We will look forward to seeing the announcement.
A human annotation tool that works well with UIMA would be an important
addition to our ecosystem, so I'm glad this discussion is happening.
BTW, are people familiar with the annotation standards work that is going
on at ISO (http://www.tc37sc4.org/)? This bears watching as well, as it
might evolve into the annotation standard that everybody's been waiting for.
At least that seems to be the plan of the folks working on it ;-).
--Thilo
