An information extraction toolkit.
To discuss the project with use, join our maing list: http://groups.google.com/forum/?fromgroups#!forum/bluestocking-dev
This project depends on NLTK. You will need to install it before running these scripts.
To run tests:
python tests.py
To run factchecker demo, try this:
python factchecker.py “The sky is not blue.”
or this:
python factchecker.py “People never eat fish. Goldfish are unpopular.”
This test a document against the Simple English Wikipedia articles for each word in the string passed as argument.
(Warning: documents with long sentences take longer to query)
Scripts included:
### parse.py
Defines Document class for wrapping raw text and Parser class for extracting Relations from a Document.
Documents have a method to turn them into Doxaments (see below).
### doxament.py
Defines a Doxament class. A Doxament contains many Relations. A Doxament may be queried for consistency with another Doxament. They may also be merged to form a more complete knowledge base.
Relations encapsulate a semantically significant lexical cooccurence.
### other
wikipedia.py and wiki2plain.py from http://stackoverflow.com/questions/4460921/extract-the-first-paragraph-from-a-wikipedia-article-python
