Table of Contents

• StarGraph
• Features
• Demo: Querying The Wikipedia Knowledge Graph
  • Software Requirements
  • Hardware Requirements
  • Installation
    • Starting Up
    • Load DBpedia-2016 data
  • Querying DBpedia

StarGraph

An Open Source graph database to query large Knowledge Graphs. StarGraph support natural language queries facilitating the interaction between domain experts and data.

Features

• Natural language interface
• Easy integration of structured and unstructured data
• Built-in semantic approximations
• Multi-lingual
• REST-like interface

Demo: Querying The Wikipedia Knowledge Graph

To give you a more hands-on experience we've prepared a docker based deployment ready to consume the public dumps of DBpedia 2016. Please follow this instructions:

Software Requirements

These are our recommendations and does not mean that Stargraph will not work on other tools and platforms. We appreciate feedbacks as well.

• Docker and Docker Compose: Our test environments are using version 1.17 and 1.11 respectively.
• Linux OS: We adopted Ubuntu LTS Server. Probably from ubuntu trusty onwards you're safe.

Hardware Requirements

These are our recommendations.

• Around 60GB of free disk space (SSD is a big plus)
• 16GB of RAM but we strongly recommend 32GB
• +4 CPUs
• Internet connection to download all resources.

Installation

Stargraph uses two majors Open Source software components:

• Indra
• Elasticsearch

For this demonstration we use a public hosted version of Indra. More information here hence you need internet connection. If this is not possible follow this instruction to have Indra in your infrastructure.

Starting Up

Docker Compose will assemble all software components and ease the setup of this Demo. Just copy and paste this file as docker-compose.yml in your filesystem.

version: '2'

services:

  coreference:
    depends_on:
      - corenlp
    image: "lambdacube/pycobalt:v1.1.1"
    environment:
      - PYCOBALT_CORENLP=http://corenlp:9000
    expose:
      - "5128"
    ports:
      - "5128:5128"

  corenlp:
    image: "lambdacube/corenlp:3.7.0"
    expose:
      - "9000"

  elastic:
    image: elasticsearch:5.1.1
    container_name: elastic
    command: "-Ecluster.name=StarGraphV2Cluster -Ebootstrap.memory_lock=true -Ediscovery.zen.minimum_master_nodes=1"
    expose:
      - 9300
    volumes:
      - ./esdata:/usr/share/elasticsearch/data
    ulimits:
      memlock:
        soft: -1
        hard: -1
      nofile:
        soft: 65536
        hard: 65536
    mem_limit: 8g
    cap_add:
      - IPC_LOCK
    environment:
      - "ES_JAVA_OPTS=-Xms7g -Xmx7g"

  stargraph:
    image: lambdacube/stargraph:latest
    container_name: stargraphv2
    ports:
      - 8917:8917
    volumes:
      - ./logs:/usr/share/stargraph/logs
      - ./data:/usr/share/stargraph/data
    mem_limit: 8g
    environment:
      - "STARGRAPH_JAVA_OPTS=-Xms7g -Xmx7g"

.. now run the following command in your favorite terminal:

$ docker-compose up

This will download the docker image for Stargraph and also setup a local instance of Elastic Search Server from the officials repositories. On the first run this can take some minutes to finish.

After a while you should notice that Stargraph is up and listening on port 8917.

Load DBpedia-2016 data

Stargraph needs to build a set of indices to be able to answer your queries. Ensure that Stargraph is listening on port 8917 and issue the following command using cURL or any other HTTP client you like. At this point Stargraph will use internet connection to fetch the DBpedia dumps.

First test. List the pre-configure datasets for DBpedia.

$ curl -XGET http://localhost:8917/_kb
["dbpedia-2016/facts","dbpedia-2016/entities","dbpedia-2016/relations"] 

It's all good. Let's populate those indices.

$ curl -XPOST http://localhost:8917/_kb/dbpedia-2016/_load
{"ack": true}

Stargraph acknowledges your command and will start building the indices. Actually this can take few hours depending on your hardware.

Querying DBpedia

Use your favorite HTTP client to ask something to Stagraph: Who is the wife of Barack Obama?

$ curl -v -XGET "http://localhost:8917/_kb/dbpedia-2016/query?q=Who%20is%20the%20wife%20of%20Barack%20Obama?"

Note that the question is URL encoded, this is a requirement to use cURL properly.

And the output is a JSON response.

{
  "query": "Who is the wife of Barack Obama?",
  "sparqlQuery": "SELECT * WHERE {\n{<http://dbpedia.org/resource/Barack_Obama> <http://dbpedia.org/ontology/child> ?VAR_1} UNION \n{<http://dbpedia.org/resource/Barack_Obama> <http://dbpedia.org/property/spouse> ?VAR_1} UNION \n{<http://dbpedia.org/resource/Barack_Obama> <http://dbpedia.org/property/children> ?VAR_1}\n}",
  "interactionMode": "NLI",
  "answers": [
    {
      "id": "dbr:Michelle_Obama",
      "value": "Michelle Obama",
      "score": 1
    },
    {
      "id": "\"1992-10-03\"^^<http://www.w3.org/2001/XMLSchema#date>",
      "value": "1992-10-03",
      "score": 1
    }
  ],
  "mappings": {
    "Barack Obama": [
      {
        "id": "http://dbpedia.org/resource/Barack_Obama",
        "value": "Barack Obama",
        "score": 1
      }
    ],
    "wife": [
      {
        "id": "http://dbpedia.org/ontology/child",
        "value": "child",
        "score": 0.560512271511479
      },
      {
        "id": "http://dbpedia.org/property/spouse",
        "value": "spouse",
        "score": 0.5078599088550189
      },
      {
        "id": "http://dbpedia.org/property/children",
        "value": "children",
        "score": 0.5803166605881719
      }
    ]
  }
}

Contributors (alphabetical order)

• Andre Freitas
• Bernhard Bermeitinger
• Leonardo Souza
• Matthias Cetto
• Siegfried Handschuh