Mission Possible: Deploying Government Linked Data (Pt1)

Sandro Hawke, (sandro@w3.org), W3C/MIT, @sandhawke
John L. Sheridan, @johnlsheridan
gov 2.0 expo, May 25-26, 2010, Washington DC
http://www.w3.org/2010/Talks/0525-linked-data (wiki)

Mission Possible

Part 1:

What, exactly, is Linked Data?
- The big picture and today's building blocks

Part 2:

Viewing Your Data as Triples (slides)
- Working with RDF vocabularies; communicating via triples

Part 3:

Publishing Triples on the Web (slides)
- The mechanics and politics of actually making the data available

Part 1: Fundamentals

@@ fill in overview of of part one, in 3-8 bullet points

About Us

@sandhawke
- programmer (C, C++, Java, Perl, Prolog, now mostly Python)
- at W3C since 2000, doing RDF, OWL, RIF, SPARQL, Govt
- W3C: consensus standards, founded by TimBL 1994
  - about 60 Working Groups doing HTML(5), CSS, SVG, XML, Accessibility

@johnlsheridan
- Civil Servant since 2004
- lead on Linked Data for data.gov.uk
- co-chair of W3C e-Government IG

About You (Just Curious)

Do you work for the government (Federal, State, City, County) or for a supplier?
Can you read/write HTML?
Can you read/write some data format:
- JSON, XML, SQL?
Can you program in some language:
- C, C++, Java?
- Perl, Python, Ruby?
- Javascript?
- XSLT?
Do you know how HTTP works:
- Response Codes (eg 403 Forbidden)?
- Content Types and Content Negotiation?
- RESTful APIs?
Formal math, logic, logic programming?
- The difference between "domain" and "range"?
- The Unique Names Assumption? Negation-as-Failure?

About the "Semantic Web"

The Web should be more than just documents for people to read
Allow machines to traverse, aggregate, analyze, answer
TimBL's vision, but it's a big tent
- "Semantics", "Ontologies" (Research Funding)
- Web Architecture, REST

See Kate Ray's Web 3.0 Video (esp. until 3:37 or 6:50)

Linked Data has a narrower goal; uses some of the same technologies.

What Is Linked Data?

Extending spreadsheets and databases to work over the Web.

Give web identifiers (URIs) to things
Publish information about them as Web Resources (good website architecture)
Use Triples (subject, property, value)

Data at http://dbpedia.org/page/Massachusetts
Subject	Property	Value
http://dbpedia.org/resource/Massachusetts	nickname	"Bay State"

Why does Linked Data make sense for government?

Responsible Publishing of data (we'll explain why)
Combine different data about the same things, although it is held by different parts and levels of government
Can make it easier for people to consume your data (we'll explain how)

RDF Triples

Quick Demo

Data at http://dbpedia.org/page/Massachusetts
Subject	Property	Value
http://dbpedia.org/resource/Massachusetts	nickname	"Bay State"

sameAs.org says:

explore with a Linked Data Browser

... and try to imagine you're a machine.

So How Do You Add Your Data?

Think in Subject-Property-Value Triples
Use URIs
Publish on the Web

URIs (A Little Web Architecture)

URIs are like URLs, with a few extra tricks.

Long history, "Web Architecture", lots of debate.

Here it is, put simply.

Information Resource

Anything whose current state can be entirely represented in bytes.
This is what we see on the Web, including:
- documents (maintained, or frozen)
- video and audio recordings
- photographs, drawings
- databases (product catalog)

Resource

Anything at all. Anything anyone can conceptualize.
Includes Information Resource, of course
Also includes:
- specific people, cities, countries
- my dog, the set of all dogs, the set of all animals
- ... even Unicorns and Dragons

art credit

URLs identify Information Resources

URLs are "Web Addresses"
Such as http://www.w3.org/People/Sandro or http://whitehouse.gov
There are standard Internet protocols (RFC 2616):
- given a URL, get the bytes which convey the current state of that information resource
- every browser (web client) does this

URIs identify Resources

They often look just like URLs (many are URLs)
But they can behave differently
- With some, there is no protocol for GET
- If it's a URL, the GET works normally
- For other, GET results in a 'redirect to an Information Resource
  - try to GET http://dbpedia.org/resource/Massachusetts
  - redirect ("303 See Other") to http://dbpedia.org/page/Massachusetts
  - which has data about http://dbpedia.org/resource/Massachusetts

Sometimes IRI when it can contain non-ascii characters, like http://www.example.org/rosé .

Hash and Slash

Two kinds of indirection:

Hash URIs contain a hash ("#") character: http://vocab.deri.ie/dcat#granularity
- chop off the hash and everything after it
- do your GET on what's left
- see what the result says about full, original URI
- Sometimes it turns out to be a "fragment" URL
  - Hopefully it's not both (but it happens sometimes)

Slash URIs don't contain a hash: http://purl.org/dc/elements/1.1/creator
- try to do a GET
- you might just get contents (so it was a URL)
- you might get a "303 See Other" redirecting you elsewhere
- Hopefully the new place says something useful using the original URI.

Hash vs Slash

Hash URIs:

easier to construct
more efficient on a small scale
often used for small, controlled situations

Slash URIs:

more control over user experience
better scaling

You'll see both.

When publishing, your software may choose for you.

Review

Use URIs to identify things
Think in Triples
Publish on the web

Publishing Data

There's lots of choice!
Don't be overwhelmed, it means there's at least one method that will work well in your situation.

Publication Method	Advantages	Disadvantages
RDF/XML Document	Oldest, best supported	Confusingly like normal XML
Turtle (N3) Document	Simplest	Not technically a standard yet
HTML Document with RDFa	Fits inside HTML attributes	Can get very complicated
JSON	Normal JSON, but also RDF	Promising, but still being developed
GRDDL	Use the XML you have/want	Needs to download+run XSLT
SPARQL	Query Protocol	Query Protocol

RDF/XML Example

<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
         xmlns:db="http://dbpedia.org/resource/">
  <rdf:Description rdf:about="http://dbpedia.org/resource/Massachusetts">
    <db:Governor>
      <rdf:Description rdf:about="http://dbpedia.org/resource/Deval_Patrick" />
    </db:Governor>
    <db:Nickname>Bay State</db:Nickname>
    <db:Capital>
      <rdf:Description rdf:about="http://dbpedia.org/resource/Boston"> 
         <db:Nickname>Beantown</db:Nickname>
      </rdf:Description>
    </db:Capital>
  </rdf:Description>
</rdf:RDF>

validator

Turtle Prefixes

First triple:

<http://dbpedia.org/resource/Massachusetts> 
   <http://dbpedia.org/resource/Governor> 
       <http://dbpedia.org/resource/Deval_Patrick> .

Abbreviate it:

@prefix db: <http://dbpedia.org/resource/>

db:Massachusetts db:Governor db:Deval_Patrick.

Read the same by turtle parsers

Turtle Example

@prefix db: <http://dbpedia.org/resource/> 

db:Massachusetts db:Governor db:Deval_Patrick;
                 db:Nickname "Bay State";
                 db:Capital db:Boston.
db:Boston        db:Nickname "Beantown".

RDFa Example

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML+RDFa 1.0//EN"
    "http://www.w3.org/MarkUp/DTD/xhtml-rdfa-1.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
      xmlns:db="http://dbpedia.org/resource/"
      version="XHTML+RDFa 1.0">
  <head>
    <title>About Massachusetts</title>
  </head>
  <body>
    <div about="http://dbpedia.org/resource/Massachusetts">The
    Massachusetts governor is
      <span rel="db:Governor">
	<span about="http://dbpedia.org/resource/Deval_Patrick">Deval
	Patrick
	</span>,
      </span>
      the nickname is "<span property="db:Nickname">Bay State</span>",
      and the capital
      <span rel="db:Capital">
	<span about="http://dbpedia.org/resource/Boston">
	  has the nickname "<span property="db:Nickname">Beantown</span>".
	</span>
      </span>
    </div>
  </body>
</html>

distiller community site

One Possible RDF-JSON Example

{ "__iri": "db:Massachusetts",
  "db:Nickname": "Bay State",
  "db:Governor": { "__iri": "db:Deval_Patrick"  },
  "db:Capital": {  "__iri": "db:Boston",
                   "db:Nickname": "Beantown"
                 },
  "__prefixes": { "db:": "http://dbpedia.org/resource/" }
}

One Possible GRDDL Example

<MyDataSet xmlns="http://example.org/my-data-xml-namespace">
  <State>
    <name>Massachusetts</name>
    <governor>Deval_Patrick</governor>
    <nickname>Bay State</nickname>
    <capital>
      <name>Boston</name>
      <nickname>Beantown</nickname>
    </capital>
  </State>
</MyDataSet>

All the hard work is done by an XSLT program downloaded via the XML namespace URL. (Not implemented for this demo, sorry.)

spec demo service

SPARQL

A query language, somewhat like SQL

prefix db: <http://dbpedia.org/resource/>
prefix dbo: <http://dbpedia.org/ontology/>
SELECT ?dnym WHERE { db:Massachusetts dbo:demonym ?dnym }

prefix db: <http://dbpedia.org/resource/>
prefix dbo: <http://dbpedia.org/ontology/>
SELECT ?cap WHERE { db:Massachusetts dbo:capital ?cap }

dbpedia sparql service and sparql tutorial

Content Negotiation

Information Resources can have multiple Representations
When you GET, you can say which type you want (HTML or XML say)
HTTP Server returns an appropriate representation

Try:

curl -L --header "Accept: application/rdf+xml" http://vocab.deri.ie/dcat

curl -L --header "Accept: text/turtle" http://vocab.deri.ie/dcat

curl -L --header "Accept: text/html" http://vocab.deri.ie/dcat