Skip to main content

An information extraction toolkit built on top of NLTK.

Project description

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

Project details


Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page