Question answering system

This repository contains all the necessary files required to run the materials of the "COVID-19: Automated NLP-based question answering" workshop.

Requirements

• docker
• docker-compose

Setting up the environment

The whole system requires to run 3 containers:

1. JupyterLab for the interactive notebooks
2. bert-as-service for hosting the BERT model server
3. bert-as-service for hosting the SciBERT model server

The process of setting up all the necessary dependencies requires to launch just a single command in the console:

docker-compose up

This may take a while. The necessary BERT-based models have to be downloaded. These heavy files are necessary to have working demo, but whole process is automated.

As a result, three necessary services are going to be launched, and JupyterLab will expose itself on http://localhost:8888. This is going to be one and the only gateway to the created system.