NLTK: NLP in Python
by Bijan Parsia
I’m planning to play around a bit with natural language processing for knowledge acquisition. (Two areas of interest to me: One is helping me, a fairly sophisticated reader, analyze and manage the web pages I read, esp. for the purpose of producing reports, aka blog posts, about a topic. My playing with Diigo is in this vein. The other is a knowledge representation based or enhanced Wikipedia. These two are, obviously, not unrelated.)
My standard first move would be to get a good textbook and start working with Prolog since 1) Prolog has a rich history of use in NLP, 2) I like Prolog, and 3) I really like my favorite Prolog, SWI-Prolog, which has a pretty nice bundled set of libraries and tools. But I just spied NLTK which seems to be an excellent, pedagogically oriented toolkit. I was quite attracted by the tutorial for the older version, but it isn’t clear if it will be updated. (Alas, the NLTK-Lite tutorial is just a set of slides. These are welcome, of course! And I shall definitely read through the older tutorial as well.)
Cypher also seems neat. At least, and as usual, in principle.
October 9th, 2006 at 3:36 am
You should also check out ConceptNet & it’s underlying Multilingua:
http://web.media.mit.edu/~hugo/conceptnet/
http://web.media.mit.edu/~hugo/montylingua/
Both are also Python-based & very impressive (based on user-generated common-sense knowledge-base).
Regarding books: you’ll see that there are 2 paradigms (logic vs. statistics). The logic-approach (such as in Allen’s book) works nicely but requires a lot of tedious maintenance. Personally I believe more in the statistics approach (such as in Manning & Schuetze’s book), which seems like the more strategically sound paradigm.
October 10th, 2006 at 3:42 am
Thanks for the links! I’ll check them out.
I’m aware of the difference and what seems to be the current preferred approach (stats).
The commonsense based one looks absolutely fascinating! And very useful for my current class. Thanks again.
October 13th, 2006 at 7:22 pm
[...] Tales of a Semantic Web Consultancy » Blog Archive » NLTK: NLP in Python (tags: hlt) [...]
November 16th, 2006 at 5:40 am
You wrote: “Alas, the NLTK-Lite tutorial is just a set of slides”—in fact it is currently 250 pages of freely downloadable textbook. Please see http://nltk.sf.net/lite/doc/en/
November 24th, 2006 at 2:00 am
Oh! thanks Steven!
June 7th, 2008 at 4:03 am
And may be you have a look at http://attempto.ifi.uzh.ch/site/ as they have CNL 2 OWL parser.