In the Java world, there is GATE (general architecture for text engineering) and it seems very impressive. Are there something like that for Python?
I worked with GATE this last summer and really hated it. Can't decide whether that was just my growing distaste for Java or actually the GATE API. Anyway, if you're looking for something like GATE that (in my experience) runs significantly faster, you should look at Ellogon (www.ellogon.org). It's written in C and TCL, with C++, Java, Perl, and Python bindings. And I believe, if you have any software already written for GATE, Ellogon can run those modules directly. I've personally never done so -- all my modules are written in Python (often simple wrappers for things like MXPOST, MXTerminator, Charniak's parser, etc.) I find the Python interface simple and easy to use, and they've added a number of my suggestions to the API in the last release.
STeVe -- http://mail.python.org/mailman/listinfo/python-list
