Hello everyone,
when I started to work on the Cas Editor a few years back we just
needed an annotation tool which could be used by a small team to annotate
a few hundred doctor letters. While having that in my mind I build it
very restricted
for that specific use case.
The move to the uima sandbox changed the situation and people also wanted
to view/edit CASes during the development of AEs.
Initially the Cas Editor was packaged as an eclipse RCP application which
made this new use case impossible. As a consequence is was turned into
an eclipse
plugin which can be installed together with the other uima plugins.
But there are still a few remaining "features" which are too specific
for the annotation
project use case. The Cas Editor project model does not work well
together with
the eclipse java project model JDT needs to organize all the java
artifacts. Actually
as far as I know there is no way to create a project with the eclipse UI
which has
the Cas Editor and JDT nature.
There are more limitations in the development use case, it is right now
not possible
to run an AE against a CAS document collection directly from the eclipse
AE java project.
The Cas Editor has basic support to run an AE, but that only works when
the AE is either
packaged as a PEAR or installed as an eclipse plugin, both ways are kind
of annoying
for that use case. The Cas Editor does not integrate well with other
uima eclipse tooling,
e.g. the editors for the descriptor files.
After the Cas Editor was turned into an eclipse plugin people also wanted
to integrate it into their own internal/proprietary eclipse tooling.
This use
case should work actually but was never really tested by me and is
undocumented,
so in theory it should work and in practice me might have to still fix a
few small bugs.
In the future I actually want to limit the Cas Editor to only view/edit
CASes and let
the other functionality be done by specific eclipse plugins. To be able
to use it
there must be an additional plugin which extends the Cas Editor and knows
how to open CASes (with the correct type system), provide configuration,
etc.
For the UIMA eclipse tooling that would mean that we need a development
plugin which organizes the CASes in a way which works nicely with JDT
and provide
some JDT launch wizard to run an AE against a document collection.
This development plugin can also be used by users who want to do basic
CAS editing
as its possible now.
People who did use the Cas Editor as part of an annotation project
either have to use it
with the uima development plugin or integrate it into their own tooling.
In the long term
I want to provide tooling infrastructure for annotation projects which
create
training data for statistical NLP tools like they can be found in OpenNLP.
Other opinions ?
Jörn
