A Question and Answering system for the DCAT vocabulary with a Social Media Bot as user interface developed as part of a bachelor thesis.

DCAT QA System

See qa-system readme

Twitter Bot

See twitter-bot readme

Botkit Twitter Adapter

The code for the Twitter adapter for botkit can be found on Github but a copy is included in the project as well under sninterface-botkit/twitter-bot/botbuilder-adapter-twitter

Web Bot

See web-bot readme

Nginx Reverse Proxy Config

See nginx config

Local Building and Deployment of all Components

Prerequisites:

Docker
Docker-Compose
maven
ngrok (only for twitter bot)
npm

Steps:

1. Build Docker Images

qa-system:

Run in dcat-qa-system:

mvn clean package -DskipTests
docker build -t qa-system .

See qa-system readme

sninterface-twitter:

Run in sninterface-botkit/twitter-bot:

docker build -t twitter-bot .

See twitter-bot readme.

sninterface-web:

Run in sninterface-botkit/web-bot:

docker build -t web-bot .

See web-bot readme

2. Copy and Fill out .env Templates

qa-system:

change SPARQL_ENDPOINT when running your own Fuseki instance (see Running your own fuseki triplestore instance)

twitter-bot (can be skipped if web interface is sufficient):

fill in (see Twitter Bot Setup for how to obtain these values)

TWITTER_CONSUMER_KEY
TWITTER_CONSUMER_SECRET
TWITTER_TOKEN
TWITTER_TOKEN_SECRET
TWITTER_WEBHOOK_ENV

start ngrok with

ngrok http 3000

and fill in the generated https address at WEBHOOK_URL (this will expose your local port 3000 to the internet so Twitter can send events to your application)

web-bot:

nothing to be done

3. Create and Start All Containers

run

docker-compose up

run

docker-compose up sninterface-web qa-system elastic

to run everything except the twitter bot

4. Ask Bot Questions via Web Interface or the Registered Twitter Account

The web interface should be available under http://localhost:3000 (unless the port was changed)

Deployment without Building

Already built docker images of the qa-system, twitter-bot and web-bot can be downloaded from the repository's registry.

Running Your Own Fuseki Triplestore Instance

If the fuseki endpoint at https://openbot.cs.upb.de/fuseki/ is not available anymore, it may be required to locally deploy a fuseki instance providing the data that should be queried by the qa-system.

The docker page of the Jena Fuseki 2 image explains how to set up a fuseki instance as docker container and import data. The docker-compose.local.yml also contains a comment with configuration for the fuseki docker, which can be used as a starting point.

The current opal graph data is available at https://hobbitdata.informatik.uni-leipzig.de/OPAL/OpalGraph/DCAT-QA/.

projekt-opal/dcat-qa