Another option for doing (some) of NLTK in UIMA is to use the BSF Annotator in the sandbox with Jython. This would only work for the pure Python components and not the stuff that relies on NumPy etc. Jython has been in better shape of late so it might be worth trying and probably better than serializing and deserializing. I do, however agree with Edward that a "native" Python interface would be nice.
Eddie, What is the status of UIMACpp right now anyway? Development on the Java side is so fast and furious (good job guys) that it is understandable that the C++ is being ignored but maybe it is a place that more community members could help? Frank -----Original Message----- From: Eddie Epstein [mailto:[EMAIL PROTECTED] Sent: Friday, February 22, 2008 12:02 PM To: [email protected] Subject: Re: Python support for reading & writing UIMA data files Enabling use of NLTK with UIMA looks like a great opportunity for the UIMA community. The SWIG'd Python enablement would presumably be able to do that, but given your comments about SWIG creating an unnatural interface, we would be crazy not to want to take advantage of your new native pycas implementation.
