Metaphor ADP Server

This repository contains a server for running abductive interpretation of metaphors. It includes a web interface and a JSON API.

Installation

Docker

The easiest way to install the server is using Docker.

1. Follow the instructions to install Docker.

2. Clone the Metaphor ADP server repository:

git clone /~https://github.com/isi-metaphor/metaphor-adp-server

3. Build the Docker image for the metaphor server:

./build

4. Get a license for Gurobi for the virtual machine. Go to https://www.gurobi.com/features/academic-named-user-license to get a key and then run

./run-license [key]

The license file will be stored in the lic directory.

5. Run the Docker image for the metaphor server:

./run

The server will now be started on localhost, port 8000. You can send JSON requests (see, e.g., the testing directory in the Metaphor ADP repository) or use the Web interface: http://localhost:8000/app

The default username and password are both metaphor.

Legacy

The directory structure should look like this:

The instructions below assume you're installing in /research, but this can be any directory you prefer.

This is how to install the entire metaphor pipeline, including the web service in this repository:

1. Clone repositories for the Metaphor ADP pipeline and for the Henry abductive reasoner:

mkdir -p /research/repo
cd /research/repo
git clone /~https://github.com/isi-metaphor/metaphor-adp.git metaphor
git clone /~https://github.com/isi-metaphor/henry-n700.git henry

2. Compile Henry:

• Install dependencies: python-dev, libsqlite3-dev, graphviz and python-lxml
• Compile:

cd /research/repo/henry
make -B

3. Install Boxer:

• Install Prolog: sudo apt-get install swi-prolog
• The official Boxer Subversion repository is no longer available, so clone our unofficial Git repository:

cd /research/repo
git clone /~https://github.com/jgordon/boxer
cd boxer
make
make bin/boxer
make bin/tokkie

• Uncompress the model in the Boxer installation directory:

cd /research/repo/boxer
tar xvjf models-1.02.tar.bz2

4. Install Gurobi: Install gurobi in a separate subdirectory of installation-dir (called gurobi maybe), and install the license file somewhere.

5. Clone the web service and set up its directories:

cd /research/rep
git clone /~https://github.com/isi-metaphor/metaphor-adp-server.git

mkdir -p /research/temp/metaphor-adp-server.tmp
mkdir -p /research/data/metaphor_kbs
mkdir -p /research/data/metaphor-adp-server
mkdir -p /research/logs/metaphor-adp-server

6. Install dependencies:

sudo apt-get install fabric openssh-server screen
sudo pip install jinja2 django lz4 gitpython pexpect regex sexpdata simplejson

7. Deploy the web service: The git clone command just downloads the code but then it needs to be configured and deployed.

• Edit the file fab/config.{branch_name}.json depending on if you want to install the prod or dev branch of the web service
• For local deployment -- that is, on the same machine:
  • Check the function install in fabfile.py to make sure it runs: basic_config(branch_name), local_config(), deploy()
  • Execute fab install:branch_name
• For remote deployment -- that is, the deployment directory is on another machine to which you have ssh access without password:
  • Edit the ssh username and ssh private key to use (remote_config function in fabfile.py)
  • Make sure you replace local_config() with remote_config() in the install function in fabfile.py
  • Execute fab install:branch_name

8. Initialize the database file:

• Go into the deployed directory and run:

python2.7 manage.py syncdb --noinput --settings=adpsrv.settings

• Create a user for the web interface by running:

python2.7 manage.py createsuperuser --username=username_to_create \
    --email=whatever@whatever --settings=adpsrv.settings

9. Run:

• Edit the shell.sh script to set the name of the screen session it'll start (the -S option)
• Run the shell.sh command
• Within the screen it started, start temp_run.sh; for monitored mode run run.sh.
• Disconnect from the screen (Ctrl-A D)

Acknowledgments

This work was supported by the Intelligence Advanced Research Projects Activity (IARPA) via Department of Defense US Army Research Lab contract W911NF-12-C-0025. The US Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright annotation thereon. Disclaimer: The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of IARPA, DoD/ARL, or the US Government.