Linked Data & the Semantic Web—Making Human Knowledge Programmable

Alexander O'Connor, @uberalex
CNGL Centre for Global Intelligent Content,
KDEG, School of Computer Science & Statistics,
Trinity College Dublin, Ireland.

This research is supported by the Science Foundation Ireland (Grant 12/CE/I2267) as part of the Centre for Next Generation Localisation at Trinity College Dublin.

"Linked Data & the Semantic Web" by Alexander O'Connor is licensed under a Creative Commons Attribution 4.0 International License.

Markup
Querying & Exploring
Hosting

45 minutes is actually not a long time to discuss the semantic web. What I'm going to try and focus on is the practical aspects in three areas:

marking up documents and designing
querying linked data online
hosting linked data

I gather most of the audience is at least familiar with the basic concepts. I will therefore only speak about them briefly. I hope to make this a pretty hands-on experience, you can follow the talk at the link above. The slides should be online, and several examples are linked. I would greatly appreciate any feedback, so be sure to hit the Github Repo for the talk! In particular I will want to cover the notions of best-practice, openness and hopefully something that you can use.

The real objective of the Linked Data Web is to make a web that is more understandable for computers. Humans communicate in text, big blocks of the stuff, that is very hard to process automatically. We use all sorts of complex linguistic tricks: metaphor, references; we often speak indirectly about things, and it all depends on the fact that we can communicate using a shared context. The goal of structuring linked data is exactly that: how do we establish a common reference model to allow computers to 'read' documents.

"My centre is giving way, my right is in retreat, situation excellent. I attack." ― Ferdinand Foch

Taking Marechal Foch's sentence as an example, what might we want to do to provide context?

Battle of the Marne & military terms
Irony
Who is Foch?
quote
translation

Background image reference

Sowa, J.F. Ontology, Metadata, and Semiotics

When humans read a piece of text, what they are doing is looking at labels which they translate into concepts. We have the concept of the Triangle of reference. There's a thing, we have a conceptualisation of that thing, and we communicate that with a symbol. There's lots of complexity there in reconciling the three.

Book Bindings

Adhesive bindings, Caoutchouc bindings, Backless bindings, Boards, Edges of binding boards, Bevelled edge boards, Square edge boards, Hemp boards, Pasteboard, Wooden boards, Mauchline ware bindings, Case bindings, Circuit edges, Dos-à-dos bindings, Flap bindings, Girdle books, Guard books, Limp bindings, Mechanical bindings, Comb bindings, Spiral bindings, Non-adhesive bindings, Tacketed bindings, Tucks, Wrappers, Cloth wrappers, Printed wrappers, Vellum wrappers, Yapp style bindings

Wrappers
broader than
Vellum Wrappers

Instances (or Individuals) are members of classes, with properties, and classes can have superclasses.

Common Understanding

Vocabulary:
Set of Terms
Taxonomy:
Terms & Concepts & Relationships
Ontology:
Terms & Concepts & Relationships & Data

Subject—Predicate—Object

	Year	City
i	1978	Kraków
ii	2005	Munich
iii	2013	Buenos Aires

This gets assembled into a graph structure, which can then be traversed algorithmically

Schema

Class: a group of instances with common properties
An individual is a class
Individuals have properties with values

OWL—Web Ontology Language

Classes can have rich semantics.
Reasoning using OWL allows creation of new knowledge.
Don't re-invent RDF:Wheel!

have a notion of semantics, but how do we actually use that to improve what we have?

http://data.dws.informatik.uni-mannheim.de/lodcloud/2014/

We can leverage the openness of the LOD web to get our hands on concepts, on data, on terms, and link it all together to get a network effect.

Select, Describe, Construct, Ask

Graph, Table, XML, JSON, Text

What to enrich and how?

add metadata, add context, add facts, resolve ambiguities, mark linguistic factors, licensing, etc etc?

PapyrusCase55_041	Title:	An Irish hero! 1 Irishman defeats 10 Germans
	Name(s):	Great Britain. Army.
	Department:	Early Printed Books, Trinity College Library, Dublin
	Item No.:	EPB Papyrus Case 55b
	Collection title:	World War I Recruiting Posters
	Is part of:	World War I Recruiting Collection
	Digital No.:	PapyrusCase55_041
	Abstract:	World War I Recruiting Collection: These posters and leaflets were issued by the British Army in Dublin during the war.
	Type of work:	posters
	Dimensions:	78 x 53 cm
	Material:	paper (fiber product) colored ink
	Subjects:	Patriotism World War, 1914-1918--Great Britain--Posters Recruiting and enlistment World War, 1914-1918--Ireland World War, 1914-1918--Campaigns--Germany
	Publisher:	James Walker
	Copyright:	Copyright 2012 The Board of Trinity College Dublin.

One thing here is that if we collect the data from the poster, he's a Sergeant. In Wikipedia, he's a Major. In the Citation, he was a Lance-Corporal.

@prefix dbpedia: http://dbpedia.org/resource
@prefix ex: http://www.example.com/

ex:Michael_John_O'Leary is a dbpedia:Military_Person.
ex:Michael_John_O'Leary ex:hasAward ex:VC3556.
ex:Michael_John_O'Leary ex:hasRank ex:Sargeant.
ex:Michael_John_O'Leary foaf:name "Michael John O'Leary".
ex:Michael_John_O'Leary OWL:sameas dbpedia:Michael_John_O'Leary.
ex:VC3556 ex:hasDate "1915-02-16^^XSD:Date".

Note, I have not validated this and it's untested, so beware.

Major Michael John O'Leary VC (29 September 1890 – 2 August 1961) was an Irish recipient of the Victoria Cross, the most prestigious award for gallantry in the face of the enemy that can be awarded to British and Commonwealth forces.

http://en.wikipedia.org/wiki/Michael_John_O%27Leary

Major
 Michael John O'Leary
 VC
 (29 September 1890
– 2 August 1961)
was an Irish recipient of the Victoria Cross, the most prestigious award for gallantry in the face of the enemy that can be awarded to British and Commonwealth forces.

 <a href="http://en.wikipedia.org/wiki/Michael_John_O%27Leary">
 http://en.wikipedia.org/wiki/Michael_John_O%27Leary
 </a>

Key question: is this complete with respect to schema, with respect to use cases?

Just get it online, whatever it is

What data do we have?

CSV
PDF
JPG
XML
TXT

Facts, Data, metadata. This is just thinking about the data level: not know information, which brings in questions of natural language, etc, or the knowledge level.

Void and LOV provide a way of making sure that you leverage the best concepts from vocabularies, and that you can document your schema to help users get the value.

Describe the dataset Interlinking statements map from your data to your semantics (eg from database, writing rdf, etc) REUSE, Interlink

Five-star Data
Endpoint Availability
Documentation
Provenance
Licensing

Openness is the key requirement for this, but it's not enough just to push content out of the nest. Open Data has to be structured in an accessible way (right metadata, right format and written in best practice) It also has to be sustained: that means using what's already out there, and being aware that you need to maintain it.