eriktks/shico-deploy

Deploying ShiCo

This repository is meant to illustrate how existing software can be used with different datasets (times, kb and german news papers), enabling scholars to carry out research on different resources.

Instructions for deploying ShiCo

We assume that you are able to run Docker images.

NOTE: on SE-Linux, make sure you can mount files -- you can run setenforce 0 (although it may lead to vulnerabilities in your system) or you change the label in the container context as shown here.

You can run the scripts provided in this repo to download the ShiCo data and deploy a ShiCo instance.

Download data for a given model

Run the download.py script:

python download.py DOI

where DOI can be:

Dataset	DOI
times	https://doi.org/10.5281/zenodo.1494140
kb	https://doi.org/10.5281/zenodo.1189327
German historic newspapers*	https://doi.org/10.5281/zenodo.3255285

Note: German newspapers contains 3 sets of chronicling_america, europeana and sbb, which can be deployed separately.

Convert to gensim

Run the convert.py script:

python convert.py

Or, for the German newspapers:

python convert.py SET_NAME

where SET_NAME can be one of: chronicling_america, europeana or sbb.

Deploy dockers

Run the start.sh bash script:

./start.sh

You should now have a ShiCo backend running on port 8000 and frontend running on port 3000 of your localhost. Visit http://localhost:3000/ to use your instance of ShiCo.