The OpenNLP Grok Library / News: Recent posts

I just realized one can indicate when a project is inactive, so I have done this now for Grok. If you are interested in the NLP tools that came with Grok, you should check out http://opennlp.sf.net. If you are interested in the CCG parsing components, see http://openccg.sf.net.

Mike is taking over Grok development and repurposing it for primary
use as a CCG Realizer in limited domain dialogue systems. See
http://www.iccs.informatics.ed.ac.uk/~mwhite/White-Baldridge-ENLG-2003-to-appear.pdf
for a description of the effort so far. Version 0.7.0 will be the
last Grok release. After this version, Grok will be split into
separately usable and separately developed OpenNLP components. Tom
Morton will be responsible for further development of the
pre-processing components. Mike will be responsible for further
development of the CCG parser and realizer.

Though I actually made the 0.6.0 release a few weeks ago, it wasn't until today that I got the file releases actually available (due to a previous ftp error). Anyway, the files are there now, and at this point I am stepping down as the main developer of Grok. As you may have seen from the previous news item, Mike White is now working to support CCG-based sentence realization in Grok, and he is taking over as the lead developer of the system overall.... read more

Now that Jason has finished his dissertation release, I'm taking over development in order to build a CCG realizer use in limited domain dialogue systems. Contact me for further details if interested.

-Mike

So, things are coming along. Most work has centered around getting the hybrid logic stuff to work with parsing. The next steps are going to be working on incorporating extensions to the CCG base. Most of the work from me will thus be focused on using Grok as a research vehicle for some time now.

This is a release in which the "deadwood" of v0.5.4 has been pulled out, hence the "Stripped" moniker. It also features important bug fixes and improvements to the preprocessing components in their handling of the new XML structure of the NLPDocument class from opennlp.common.xml.

From the release notes:
This release has come about for two reasons. First there
have been significant changes to Grok, as can be observed
in the change log. However, more importantly, there is
going to be a lot of old code ripped out of Grok in the
weeks to come, so it seemed a good idea to make a code
capsule release before doing that. This release has thus
been named the "Deadwood" release, but it nonetheless does
contain a lot of low-level improvements as well, and
reintroduces parsing functionality.... read more

The previous release had a problem in the build process that was corrupting the maxent models. This release fixes that so that the jar created by the build will behave properly for components that use maxent models.

Many improvements to the preprocessing components are now available (sponsored by eTranslate and Electric Knowledge). Parsing is broken at the moment, but that will change soon!

We are now officially using the LGPL License for Grok. The Sourceforge Trove categorization and all source files have been updated to reflect this.