hooray for Haskell NLP activity!

[Rogan Creswick]

On Fri, Nov 26, 2010 at 4:37 AM, David MacIver <david at drmaciver.com> wrote:
>
> For what it's worth, I'd love to see that. I've been keeping half an
> eye out for a good maxent package that isn't in C++ or on the JVM for
> a while (I have a project currently built in JRuby on top of
> OpenNLP/maxent which I'd love to move off the JVM and would be very
> happy to move out of ruby, but every attempt I've tried has resulted
> in the prediction quality being completely rubbish)

I'm in the same boat, more or less.  I've used the OpenNLP tools quite
a bit, and I miss them on haskell ;).

jvm-bridge looks like it might be a viable stop-gap, but native
Haskell (or ffi to c) would be *much* nicer :)

I would also like to see something along the lines of UIMA (the
Unstructured Information Management Infrastructure [1]) in Haskell,
and that might help to provide a unifying set of types / APIs to
program too.  I'm on the fence about trying for UIMA compatibility --
some of the decisions that project has made clash with my design
preferences, but I've found the general architecture to be pretty
nice.  It also isn't *technically* NLP-specific, but I've found it to
be very useful for NLP.  Particularly when experimenting with
different combinations of algorithms to achieve more complex
processing.

[1] http://uima.apache.org/

--Rogan