Local web interface to understand inference over custom knowledgebases.

Here's how to get the software, setup an existing or custom KB then access the interface.

Make you sure have virtualenv, npm and node installed.

Additionally, I've found it easier to install numpy and scipy globally and include system site packages in my virtualenv.

Here's my current setup on Mac OS X 10.9.2 for reference.

$ npm --version
1.2.17
$ virtualenv --version
1.10.1
$ node --version
v0.10.3
$ python -c "import numpy; print numpy.version.version"
1.8.1
$ python -c "import scipy; print scipy.version.version"
0.14.0rc1

Download this repo and install its dependencies.

# get the code
git clone --recursive https://github.com/jdhenke/alar.git
cd alar

# install npm dependencies
npm install
(cd celestrium && npm install)

# install python dependencies
virtualenv env --system-site-packages
source env/bin/activate
pip install pecan
python setup.py develop

To use your own KB, put it in a CSV file with each row being 3 cells of concept, relation, concept. The 4th cell can be an optional truth, which defaults to 1.0. You can get C4's assertion list with the following.

curl http://web.mit.edu/jdhenke/www/alar/assertions.csv > assertions.csv

Then, to prep your KB, run the following without virtualenv enabled:

python prep_kb.py assertions.csv

TODO: This CSV was generated using get-c4-assertions repo.

Once your KB has been prepped, run the following from inside your repo:

source env/bin/activate
. driver.sh

Now go to http://localhost:8080/.

See this illustrated tutorial of Alar.