I'm now maintaining Patrick Ye's Python library for accessing FrameNet data. This version works with FrameNet version 1.3.

First you must obtain and install the FrameNet XML files, which are free for non-commerical and require an application. Then, you can obtain the Python Library and run:

python setup.py build
sudo python setup.py install

Second, you should set your FRAMENET_HOME environment variable to point to where you installed and uncompressed FrameNet's XML files. (Alternatively you can edit the FRAMENET_PATH constant in framenet/settings.py).

Then you can test your installation by running python test/importer.py or by opening a Python shell and typing:

import framenet
fn = framenet.FrameNet()

The first time you instantiate the FrameNet class, it will parse all of the XML files and save pickle objects. You should see:

Loading the frames ... Framenet not initialized, doing it now
Loading xml for frames ...

The first time you do this it will take a while, so go have a coffee or celebrate your next birthday.

• FrameNet.lookupLexicalUnit(text, part_of_speech)

You can look up frames by the lexical units (e.g., words) and their parts of speech. For example:

buy = fn.lookupLexicalUnit('buy','v')

• FrameNet.lookupFrame(frame)

Also, you can look up FrameNet frames by their name:

commerce_frame = fn.lookupFrame('Commerce_buy')