STC Summit 2010: Semantic Web and Content Strategy

© 2010 Razorfish. All rights reserved. Confidential and proprietary.

THERE’S NO SEMANTIC WEB WITHOUT CONTENT AND DATAMAY 5, 2010

TECHNICAL COMMUNICATION SUMMIT ’10

RACHEL LOVINGER

@RLOVINGER


“Language is magic, and computers are still dumb."

- Aaron Straup Cope (flickr.com)

2


BLACKBERRY


BLACKBERRY

Photo by enrique dans

http://www.flickr.com/photos/zandperl/96744534/�

http://www.flickr.com/photos/kingdatoro/535286146/�


BLACKBERRY

Photo by Rob MacEwen

http://www.flickr.com/photos/torontorob/3818415115/�



AGENDA

‣ What is the semantic web?‣ The key ingredients‣ How it’s being used now‣ What it means for Content Strategy

6


WHAT IS THE SEMANTIC WEB?


TRANSLATE THAT INTO COMPUTER-ESE

The underlying strategy of the Semantic Web is to create data and websites that are “machine-readable.”

If machines comprehend the meaning of data and content, they can: ‣ manipulate data in more meaningful ways‣ provide precisely the information that the user wants

8


IS THERE A STARBUCKS NEARBY?

9


A FRENCH RESTAURANT?

10

© 2010 Razorfish. All rights reserved. Confidential and proprietary.11

GIFT FOR YOUR SUPERHERO NIECE?

?

? ??

??

?

?

Photo by Brendan Riley

http://www.flickr.com/photos/digitalsextant/4079601494/�

http://www.flickr.com/photos/digitalsextant/4079601494/�



FIND A HAIR APPOINTMENTSearch for specific criteria:• Highly-rated salon• Near the office• Available time that fits

your busy schedule

12


SOLVING FOR COMPLEXITY

Machines are good at complex things that people do poorly

• Computing or recalling long strings of numbers• Comparing large sets of data• Searching through millions of pages or data records for a

specific item

13Image by Eric Dobbs

http://eric22222.wordpress.com/2007/02/27/pi/�

http://eric22222.wordpress.com/2007/02/27/pi/�



SOLVING FOR COMPLEXITY

People are good at some complex things that machines don’t handle well

14

Equivalence 6:00pm and 18:00

Lumping similar things 6:00pm and 8:23am

Splitting different things 6:07:10 and 060710

Semantic systems are designed to capture the logic that will allow them to understand these types of relationships within data and use them to create new facts about the data.


THE KEY INGREDIENTS


HOW DO MACHINES KNOW WHAT DATA MEANS?

Identity + Definition + Structure

16


IDENTITY + DEFINITION + STRUCTURE

17


IDs‣ Machines need a unique, consistent way to identify a thing or concept. ‣ People can usually tell by context, but a machine needs a unique identifier to

be able to make connections or distinctions.


18

Bill Clinton = President William Jefferson Clinton

President Bush(George H. W.)

President Bush (George W.)

http://upload.wikimedia.org/wikipedia/commons/d/d4/George-W-Bush.jpeg�


IDENTITY: STANDARDS

Standard identifiers

ISBN: International Standard Book Number

ISMN: MusicISAN: Audiovisual works

19


IDENTITY: OPEN SOURCE

MusicBrainz: database of music metadata, licensed by BBC to augment web pages

The Police MBID: 9e0e2b01-41db-4008-bd8b-988977d6019a

20



21



OntologyDefine classifications, properties, relationships, and logic

Blackberry1 is a type of FruitA Fruit is an Edible Thing

Blackberry2 is a type of Wireless E-mail DeviceA Wireless E-mail Device is a Mobile Electronic Device

Properties of Edible Things:Seasonal – Yes/NoCalories – #Ingredients (optional) – other Edible Things

A Mobile Electronic Device can never be an Edible Thing.

22



23



Some non-standard ways to express semantics‣ MicroFormats – uses XHTML & HTML markup to embed meaning in a webpage

‣ hCard for contact information‣ hCalendar for events

‣ Machine Tags – definition added to simple user tagging (“folksonomy”)‣ flora:tree=coniferous‣ upcoming:event=81334

24

 This presentation was givenon 2010-04-16at the Content Strategy Forumin Paris, France.





25



26



New Web StandardsDeveloped specifically for expressing metadata and metadata relationships‣ Dublin Core – an ISO standard defining 15 common metadata elements‣ RDF – a model for expressing metadata as triples (subject-predicate-object)‣ OWL – adds semantic meaning‣ SKOS – expresses structured controlled vocabularies, taxonomies

27

Subject

Object


DUBLIN CORE METADATA INITIATIVE

28


DUBLIN CORE

A very flexible standard that defines 15 core metadata elements.‣ contributor – An entity responsible for making contributions to the resource.‣ coverage – The spatial or temporal topic of the resource, the spatial

applicability of the resource, or the jurisdiction under which the resource is relevant.

‣ creator – An entity primarily responsible for making the resource.‣ date – A point or period of time associated with an event in the lifecycle of the

resource.‣ description – An account of the resource.‣ format – The file format, physical medium, or dimensions of the resource.‣ identifier – An unambiguous reference to the resource within a given context.

29


DUBLIN CORE CONTINUED

‣ language – A language of the resource.‣ publisher – An entity responsible for making the resource available.‣ relation – A related resource.‣ rights – Information about rights held in and over the resource.‣ source – A related resource from which the described resource is derived.‣ subject – The topic of the resource.‣ title – A name given to the resource.‣ type – The nature or genre of the resource.

30


RDF – RESOURCE DESCRIPTION FRAMEWORK

31


RDF

Purpose: To provide a structure (aka framework) for describing identified things (aka resources)

Identified? The thing you’re talking about must be identified in a unique way.

http://www.foaf.com/Person#RachelLovinger

http://www.allmovie.com/Actor#WillSmith

Note: URIs (uniform resource identifiers) look like URLs, but might not represent an actual web page.

32


RDF

Composed of three basic elements‣ Resources – the things being described‣ Properties – the relationships between things‣ Classes – the buckets used to group the things

33


RDF

The elements are combined to make simple statements in the form of Triples

<Subject> <Predicate> <Object>

34


RDF


Subject

Object

35


RDF


<Subject> <Predicate> <Object>

Example statement: “Men In Black stars Will Smith”

Example triple:

<MenInBlack> <hasStar> <WillSmith>

36


RDF

Information Expressed in Triples<http://www.w3.org/2001/sw/RDFCore/ntriples/> <dc:creator> "Dave Beckett" .

<http://www.w3.org/2001/sw/RDFCore/ntriples/> <dc:creator> "Art Barstow" .

<http://www.w3.org/2001/sw/RDFCore/ntriples/> <dc:publisher> <http://www.w3.org/> .

Can also be expressed as XML<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"

xmlns:dc="http://purl.org/dc/elements/1.1/">

<rdf:Description rdf:about="http://www.w3.org/2001/sw/RDFCore/ntriples/"> <dc:creator>Art Barstow</dc:creator>

<dc:creator>Dave Beckett</dc:creator>

<dc:publisher rdf:resource="http://www.w3.org/"/>

</rdf:Description>

</rdf:RDF>

37


A SAMPLE OF RDF PROPERTIES

‣ type ‣ subClassOf‣ subPropertyOf‣ range ‣ domain‣ label‣ comment

38


RDF PROPERTIES

type – indicates that a resource belongs to a certain class

<WillSmith> <type> <Actor>

This defines which properties will be relevant to Will Smith.

39


RDF PROPERTIES

subClassOf – a class belongs to a parent class

<Actor> <subClassOf> <Person>

This means that all members of the Actor class are also members of the Person class. All properties are inherited, and new properties specific to Actor can be added.

<WillSmith> <type> <Actor>

Implies: <WillSmith> <type> <Person>

40


RDF PROPERTIES

subPropertyOf – a property has a parent property

<hasStar> <subPropertyOf> <hasActor>

This means that, if you make a statement using the hasStar property, a more general statement using the hasActor property is also true.


Implies: <MenInBlack> <hasActor> <WillSmith>

41


RDF PROPERTIES

range & domain – the types of resources that use a property

<hasStar> <range> <Actor>

<hasStar> <domain> <Movie>

This means that, if you make a statement using the hasStar property, the system will assume that the subject is a Movie and the object is an Actor.

<WillSmith> <hasStar> <MenInBlack>

is an untrue statement, but not invalid

42


RDF PROPERTIES

label – a human-readable name for a resource

<http://www.allmovie.com/Actor#WillSmith> <label> <Will Smith>

43


RDF PROPERTIES

comment – a human-readable description

<http://www.pretendwebsite.com/rlovinger-semantic-web.pdf> <comment> <A presentation that Rachel gave at the Technical Communication Summit ‘10>

44


RDF: WEB OF TRIPLES

Blackberry1

Fruit

BerryPie

EdibleThing

Blackberry1

45


RDF: WEB OF TRIPLES

Blackberry1

Fruit

BerryPie

EdibleThing

xyzabc 123In

gred

ient

Of

46


RDF: CONCLUSION

Why is RDF uniquely suited to expressing data and data relationships?

‣ More flexible – data relationships can be explored from all angles

‣ More efficient – large scale, data can be read more quickly‣ not linear like a traditional database‣ not hierarchical like XML

47


OWL – WEB ONTOLOGY LANGUAGE

48

© 2010 Razorfish. All rights reserved. Confidential and proprietary.49Winnie the Pooh and characters © A. A. Milne,, drawing by Ernest H. Shepard


OWL

Purpose: To develop ontologies that are compatible with the World Wide Web.

Ontologies? Definition and classification of concepts and entities, and the relationships between them.

50


OWL

Based on the basic elements of RDF; adds more vocabulary for describing properties and classes. Allows the creation of rules that help further explain what things mean.

‣ Relationships between classes‣ Equality‣ Richer properties‣ Class property restrictions

51


OWL: RELATIONSHIPS BETWEEN CLASSES

‣ disjointWith – resources belonging to one class cannot belong to the other

<Person> <disjointWith> <Country>

‣ complementOf – the members of one class are all the resources that do not belong to the other

<InanimateThings> <complementOf> <LivingThings>

52


OWL: EQUALITY

‣ sameAs – indicates that two resources actually refer to the same real-world thing or concept

<wills> <sameAs> <wismith>

‣ equivalentClass – indicates that two classes have the same set of members<CoopBoardMembers> <equivalentClass> <CoopResidents>

53


OWL: RICHER PROPERTIES

‣ Symmetric – a relationship between A and B is also true between B and A

<WillSmith> <marriedTo> <JadaPinkettSmith>

implies: <JadaPinkettSmith> <marriedTo> <WillSmith>

‣ Transitive – a relationship between A and B and between B and C is also true between A and C

<piston> <isPartOf> <engine>

<engine> <isPartOf> <automobile>

implies: <piston> <isPartOf> <automobile>

54


OWL: RICHER PROPERTIES CONTINUED

‣ inverseOf – a relationship of type X between A and B implies a relationship of type Y between B and A

<starsIn> <inverseOf> <hasStar>


implies: <WillSmith> <starsIn> <MenInBlack>

55


OWL: CLASS PROPERTY RESTRICTIONS

Define the members of a class based on their properties

‣ allValuesFrom – resources with properties that only have values that meet this criteria‣ Example: Property: hasParents, allValuesFrom: Human ‣ Resources that meet this criteria can be defined as also being members of

the Human class

‣ someValuesFrom – resources with properties that have at least one value that meets criteria‣ Example: Property: hasGraduated, someValuesFrom: College ‣ Resources that meet this criteria can be defined as being members of the

CollegeGraduates class

56


OWL: WEB OF TRIPLES PLUS LOGIC

Note: Blackberry2 cannot be an ingredient of BerryPie, because it’s not an EdibleThing and all ingredients of EdibleThings must also be EdibleThings

Blackberry1

Fruit

BerryPie

EdibleThing

xyzabc 123In

gred

ient

Of

57


THIS SEEMS COMPLICATED. WHY DO IT?

These capabilities allow systems to express and make sense of first order logic.

All men are mortal Socrates is a man Therefore, Socrates is mortal

58


OWL: INFERENCING

‣ Create new triples based on existing triples‣ Deduce new facts based on the stated facts

<piston> <isPartOf> <engine>

<engine> <isPartOf> <automobile>

implies: <piston> <isPartOf> <automobile>

59


OWL: THREE FLAVORS

‣ OWL Lite – uses a subset of the capabilities‣ OWL DL – uses all the capabilities, but some are used in restricted ways‣ OWL Full – unrestricted use of capabilities; no guarantee that all resulting

statements are valid

60


SKOS – SIMPLE KNOWLEDGE ORGANIZATION SYSTEM

61


SKOS

‣ Also based on RDF‣ Designed specifically to express information that’s more hierarchical – broader

terms, narrower terms, preferred terms and other thesaurus-like relationships‣ Extendable into OWL, if needed

62


LINKED DATA

63


LINKED DATA: A DISTRIBUTED APPROACH

A Web of Data

64

Image by Richard Cyganiak and Anja Jentzsch

http://richard.cyganiak.de/2007/10/lod/�

http://richard.cyganiak.de/2007/10/lod/�



LINKED DATA: A DISTRIBUTED APPROACH

One page per concept ‣ URL is a type of ID‣ “topic pages” – a powerful tool

and reference point‣ high SEO value‣ aggregate content‣ contain related data & IDs

65


HOW IT’S BEING USED NOW


WALL STREET JOURNAL MOVIE REVIEWS

67


ENRICHED SEARCH RESULTS

68

Google Rich SnippetsYahoo! SearchMonkey

+


ENRICHED VANITY SEARCH

69


GETGLUE – RATINGS AND RECOMMENDATIONS

70


GETGLUE – RATINGS AND RECOMMENDATIONS

71


THE NEW YORK TIMES – ALUMNI IN THE NEWS

72


BBC MUSIC BETA – ARTISTS PAGES

73


BBC PROGRAMME PAGES

74


DATA.GOV

“The purpose of Data.gov is to increase public access to high value, machine readable datasets generated by the Executive Branch of the Federal Government.”

75


FLUVIEWNational Flu Activity Map – a widget by CDC.gov

76


DATA.GOV.UK

77


DATA.GOV.UK APPS

Help you find things‣ A post box‣ A school‣ An affordable place to live‣ A job‣ A volunteering opportunity‣ A dentist‣ A pharmacy‣ A bike route‣ A hospital ‣ A parking spot‣ A care home

78

Cyclestreets.net


PARKOPEDIA

79


DATA.GOV.UK APPS

Get information on ‣ How taxes are spent‣ Technology investments‣ Crime stats‣ The geological makeup of your area‣ Geographical details‣ Local issues ‣ Local government‣ Health‣ Obesity‣ Real Estate

80

‣ Renewable energy projects‣ Planning Alerts‣ Anti-social behavior in the area‣ Hazardous street conditions

fillthathole.org.uk


ASBOROMETER

81


WHAT IT MEANS FOR CONTENT STRATEGY

© 2010 Razorfish. All rights reserved. Confidential and proprietary.83Photo by Jon Higgins

http://www.flickr.com/photos/jon_higgins/4345009714/�

http://www.flickr.com/photos/jon_higgins/4345009714/�



SEMANTIC CAPABILITIES

Content Strategists should get familiar with these new kinds of tools and services‣ Related Content Services‣ Advanced Media Monitoring‣ Semantic Publishing Tools‣ Semantic Ad Targeting‣ Rich Data Services‣ Machine-Assisted Tagging‣ Semantic SEO

84


RELATED CONTENT SERVICES

85

‣ Enhance existing pages‣ Identify key concepts‣ Place assets and information

on the page or link to relevant offsite content

‣ Video, images, user-generated reviews, tweets, Wikipedia entries, etc.


RELATED CONTENT SERVICES

Example Services

86

Apture Provides additional contextual information in multimedia pop-ups, drawn from places such as Wikipedia, YouTube and Flickr.

Evri Allows readers to browse articles, images, and videos related to the topic of an article or content element, and provides widgets for sidebars, posts and popovers.

Headup Provides contextually relevant material from social networks and web services.

NewsCred Augments content with related stories from 6000 top news sources, as well as topic pages and license-free photos.

Zemanta Suggests related content and pictures that editors can embed in articles or blog posts.

http://www.apture.com/�

http://www.evri.com/�

http://headup.com/about.php�

http://platform.newscred.com/�

http://www.zemanta.com/�


ADVANCED MEDIA MONITORING

‣ Track Twitter, social networks, blogs, discussion boards, content sites

‣ Track a brand, industry, domain or topic

‣ With semantic capabilities:‣ more accurate relevance‣ sentiment analysis

‣ Track ongoing stories and audience reaction

87Screenshot © 2010 Phase 2 Technology


ADVANCED MEDIA MONITORING

Example Services

88

Imooty Tracks keywords and mentions of a brand, using a simple dashboard or by creating alerts, widgets, or RSS feeds.

Inbenta Follow the topics that people in your business are following.

Lexalytics Scans what’s being said in blogs, tweets and social media to provide sentiment analysis about companies, topics and current events.

Tattler Mines news, websites, blogs, multimedia sites, and social media to find mentions of topics or issues of interest to you.

http://www.imooty.eu/�

http://www.inbenta.com/�

http://www.lexalytics.com/�

http://tattlerapp.com/�


SEMANTIC PUBLISHING TOOLS

‣ Content management tools that incorporate a wide range of structure and metadata capabilities

‣ Create and publish content encoded with semantic markup and meaningful metadata

‣ Not necessary to understand all the underlying code

‣ Streamlines the publishing process‣ Makes it faster, easier, and

cheaper to bring new content products to market

89Screenshot © 2010 Thomson Reuters


SEMANTIC PUBLISHING TOOLS

Example Services

90

OpenPublish A version of Drupal with OpenCalais machine assisted tagging and RDFa formatting built in.

Jiglu Insight Finds hidden relationships to other content you’ve published and automatically creates links.

http://drupal.org/project/openpublish�

http://www.jiglu.com/jiglu-insight/�


SEMANTIC AD TARGETING

91

‣ Analyzing content pages for message, context, or mood, and inserts relevant ads

‣ Creates highly desirable ad inventory‣ Audience targeting, without the

privacy concerns of behavioral targeting

‣ Brand protection against unfortunate term-matching

An example of non-semantic contextual ads


SEMANTIC AD TARGETING

Example Services

92

ad pepper Provides ad placement, lead generation and brand protection through semantic analysis of page content and user behavior.

Peer39 Understands the meaning and sentiment of web pages so that ads can be targeted to appropriate audiences, and also protects advertisers from having their campaigns placed on negative or objectionable content. Identifies hot topics on the fly, and quickly adapts to create new “premium” inventory.

Proximic Performs real-time content analysis to accurately target ads, builds user profiles for better audience targeting, and includes brand protection measures.

http://www.adpepper.com/�

http://www.peer39.com/�

http://www.proximic.com/�


RICH DATA SERVICES

93

‣ Enhance content with linked data

‣ Import additional information, assets, services, and user-generated content

‣ Improve SEO‣ Obtain additional data and

content for application development

‣ Data set may already include map to other desirable data and services


RICH DATA SERVICES

Example Services

94

Factual An open data platform providing tools to enable anyone to contribute and use sources of structured data.

Freebase An open, semantically enhanced database of information, similar to Wikipedia, but with structured data on millions of topics in dozens of domains.

iGlue A community editable database containing images, video, individuals, institutions, and geographic locations.

http://www.factual.com/about�

http://www.freebase.com/�

http://www.iglue.com/�


MACHINE-ASSISTED TAGGING

‣ Streamlines the process of tagging content by extracting concepts on a page‣ Suggests a set of consistent tags for each piece of content‣ Content producer approves or rejects each suggested tag

95Screenshot © 2010 Thomson Reuters


MACHINE-ASSISTED TAGGING

Example Services

96

OpenCalais Automatically tags people, places, companies, facts and events found in the content.

TextWise Generates weighted, relevant metadata based on key concepts found in the text of a document or web page.

Tagaroo An OpenCalais plug-in for WordPress.

http://www.opencalais.com/�

http://www.textwise.com/�

http://tagaroo.opencalais.com/�


SEMANTIC SEO

‣ Adds semantic markup to the content, or validates existing markup

‣ Submits it to search engines‣ Boosts search rankings‣ Makes pages more accessible for

visually impaired users‣ Displays additional business data,

content, or product information directly in search results

97Screenshot © 2010 Dapper


SEMANTIC SEO

Example Services

98

Google Rich Snippets Testing Tool

Tests webpage markup to ensure that Google’s Rich Snippets feature can interpret it correctly.

Inbenta Assists in the creation of content using the terminology of popular search queries.

Semantify(by Dapper)

Provides automated semantic enhancement of a site without changing its pages. Search engines see the site with RDFa tagging embedded in the page.

http://knol.google.com/k/google-rich-snippets-tips-and-tricks�



http://www.inbenta.com/�

http://www.dapper.net/semantify/�


ADDITIONAL RESOURCES

‣ RDF Primer (http://www.w3.org/TR/REC-rdf-syntax)‣ OWL / Semantic Web (http://www.w3.org/2004/OWL)‣ SKOS (http://www.w3.org/2004/02/skos)‣ Dublin Core (http://dublincore.org)‣ LinkedData.org – Resources from across the Linked Data community‣ Sindice (http://sindice.com) – The semantic web index‣ SchemaWeb (http://www.schemaweb.info) – A directory of RDF schemas‣ Semantic Universe (http://www.semanticuniverse.com) – Educating the World

About Semantic Technologies and Applications) ‣ Semanticweb.org – A wiki for the semantic community‣ ReadWriteWeb: Semantic Web Archives

(http://www.readwriteweb.com/archives/semantic-web/)‣ Nimble – A Razorfish-Semantic Universe report coming out soon

99


CONCLUSION

‣ Content Strategy will still be needed to help implement and use these tools‣ Related Content Services, Semantic Ad Targeting, Rich Data Services,

Semantic SEO, Taxonomy/Ontology/Controlled Vocabularies‣ Establish business rules‣ Help configure the tools‣ Periodically monitor the results‣ Make adjustments as needed

‣ Advanced Media Monitoring, Semantic Publishing Tools, Machine-Assisted Tagging‣ Ongoing interaction by insightful, skilled users‣ CS might be the primary user‣ CS might train others to get the best results from their use

100


QUESTIONS?

[email protected]: @rlovinger

http://scattergather.razorfish.com

Thank you!

101

STC Summit 2010: Semantic Web and Content Strategy

Technology

data means

data records

meaning of data

semantic web

large sets of data

complexity machines

semantic systems

content strategy