1 of 26

Progressive Generation of Knowledge Graphs

2 of 26

How Do You Get Going?

Download, install, and configure the following:

  • Virtuoso “Sponger” Module — providing Data Extract, Transformation, and Load (ETL) functionality
  • OpenLink Data Explorer (ODE) Browser Extension — providing generation of Sponger Web Service URLs
  • OpenLink Structured Data Sniffer — providing generation of Sponger Web Service URLs and/or SuperLinks from HTML document content generator using various Natural Language Processing (NLP) and Named Entity Recognition & Extraction (NER) services

License CC-BY-SA 4.0 (International)

3 of 26

What’s the Problem?

Bearing in mind the nature of a Knowledge Graph, �how could it be seeded and enriched progressively, without writing a line of code?

License CC-BY-SA 4.0 (International)

4 of 26

Why is this Important?

As an individual or a team member, your competitive advantage is inextricably linked to the production and consumption of information and knowledge created by the act of putting data in context.

License CC-BY-SA 4.0 (International)

5 of 26

How is this Achieved?

Courtesy of Virtuoso’s Sponger Middleware Module, this process is reduced to an interaction similar in flow to bookmarking — where interesting documents are tagged, based on interest, for future recall, as and when discovered while browsing.

License CC-BY-SA 4.0 (International)

6 of 26

Showcase Example

7 of 26

Document about

Edsger Dijkstra

8 of 26

Using the OpenLink Structured Data Sniffer (OSDS) - 1

Starting with an HTML document about Edsger Dijkstra, the slides that follow demonstrate how OSDS triggers the following:

  • Natural Language Processing (NLP)
  • Named Entity Recognition & Extraction (NER)
  • Identification of Extracted Entities (by looking up DBpedia, Wikidata, and other Knowledge Graphs)
  • Text Matching, Highlighting, and Hyperlinking
  • Knowledge Graph Update

License CC-BY-SA 4.0 (International)

9 of 26

Using the OpenLink Data Explorer (ODE) - 1

Having stumbled upon an article about Edsger Dijkstra on the Web, simply perform the following steps:

  1. Click on the ODE icon in your Browser’s Toolbar which returns a Description of the HTML doc presented in Entity-Attribute-Value based Property Sheet Form

License CC-BY-SA 4.0 (International)

10 of 26

Using the OpenLink Data Explorer (ODE) - 2

  • Use the “Formats” menu to select alternative Entity Description Document Types containing the same Entity Description

License CC-BY-SA 4.0 (International)

11 of 26

Using the OpenLink Data Explorer (ODE) - 3

  • You can also select the DBpedia Identifier (a Hyperlink) that denotes Dijkstra by clicking on the value of the schema:mentions property to reveal an Entity Description retrieved from the DBpedia Knowledge Graph.

License CC-BY-SA 4.0 (International)

12 of 26

Using the OpenLink Faceted Browser - 1

Alternatively, you could select the “Faceted Browser” option from the “Browse using” menu which reveals another Entity Description Page that includes additional functionality enabling deeper exploration.

License CC-BY-SA 4.0 (International)

13 of 26

Using the OpenLink Faceted Browser - 2

Presented as one of the values of the schema:mentions property is the DBpedia Identifier (a Hyperlink) that denotes Dijkstra.

Clicking on that link takes you to a detailed Entity Description page.

License CC-BY-SA 4.0 (International)

14 of 26

Using the OpenLink Faceted Browser - 3

This detailed Entity Description Page, comprising information from DBpedia, includes a rich collection of additional properties that inform deeper exploration across a variety of dimensions such as:

  • Entity Types
  • Entity Attributes

License CC-BY-SA 4.0 (International)

15 of 26

Using the OpenLink Structured Data Sniffer (OSDS) - 2

Starting with the same HTML document about Edsger Dijkstra, the slides that follow demonstrate how OSDS triggers the following:

  • Natural Language Processing (NLP)
  • Named Entity Recognition & Extraction (NER)
  • Identification of Extracted Entities (by looking up DBpedia, Wikidata, and other Knowledge Graphs)
  • Text Matching, Highlighting, and Hyperlinking

License CC-BY-SA 4.0 (International)

16 of 26

Using the OpenLink Structured Data Sniffer (OSDS) - 2

Effect of clicking on the OSDS icon installed to your browser’s toolbar.

License CC-BY-SA 4.0 (International)

17 of 26

Using the OpenLink Structured Data Sniffer (OSDS) - 3

Effect of Natural Language Processing (NLP), Named Entity Recognition & Extraction (NER), Text Matching, Highlighting, and Hyperlinking — producing Super Links

Note: All of this happens in response to a simple click on the OSDS icon installed to your browser’s toolbar.

License CC-BY-SA 4.0 (International)

18 of 26

Using the OpenLink Structured Data Sniffer (OSDS) - 4

Effect of clicking on a highlighted word or phrase: a window opens comprising automatically assigned hyperlink-based identifiers, sourced from both DBpedia and Wikidata Knowledge Graphs.

License CC-BY-SA 4.0 (International)

19 of 26

Screencast Demo

Here’s a recap of the concepts described in this presentation, via a screencast published to YouTube.

License CC-BY-SA 4.0 (International)

20 of 26

OpenLink Structured Data Sniffer (OSDS) - Query Data using SPARQL

Starting with the same HTML document about Edsger Dijkstra, you can also populate a Knowledge Graph using SPARQL by

  • Clicking on the LOD Cloud Lookup button
  • Viewing SPARQL Query Solution Page for basic insights
  • Editing an Existing Query to produce alternative insights
  • Drilling-down for deeper or newer insights using Faceted Browsing
  • Sharing Query Results Pages or specific Faceted Browsing Pages with others

License CC-BY-SA 4.0 (International)

21 of 26

OpenLink Structured Data Sniffer (OSDS) - SPARQL Query Execution & Solution

Clicking on the LOD Connectivity button triggers a SPARQL query that uses the current Document (via its URL) as the Data Source Name from which a Query Solution Page is generated (and made available in a variety of formats).

License CC-BY-SA 4.0 (International)

22 of 26

OpenLink Structured Data Sniffer (OSDS) - Query Editing 1

Effect of clicking on the Cloud & Magnifying Glass icon: a SPARQL Query Editor and URL parameters are revealed, enabling creation of new Query Solutions and Insights.

License CC-BY-SA 4.0 (International)

23 of 26

OpenLink Structured Data Sniffer (OSDS) - Query Editing 2

Effect of going into the SPARQL Query Editor in Full View.

License CC-BY-SA 4.0 (International)

24 of 26

OpenLink Structured Data Sniffer (OSDS) - Query Editing 3

Creating a new query to replace the original.

License CC-BY-SA 4.0 (International)

25 of 26

OpenLink Structured Data Sniffer (OSDS) - Query Editing 4

A new query solution page is produced by running the new query.

Clicking on any record of interest from the results table starts a Drill-Down sequence.

License CC-BY-SA 4.0 (International)

26 of 26

Drill-Down and Faceted Browsing

Select an item of interest (identified by a hyperlink) that’s also a conduit to further drill-down exploration.

This approach to exploration is commonly known as the “Follow Your Nose” pattern.

License CC-BY-SA 4.0 (International)