Building Interoperable Metadata

Size: px

Start display at page:

Download "Building Interoperable Metadata"

Pauline Norman
6 years ago
Views:

1 Building Interoperable Metadata Hal Abelson, Ben Adida MIT Computer Science & Artificial Intelligence Laboratory 8 March

2 the important thing about the semantic web is the web part. paraphrased from Chris Welty, Dan Connolly, Hal Abelson. 2

3 "If you think of semantic web as an interoperability technology, then it becomes capable of serendipitous behavior, of opportunism." Ora Lassila, yesterday. 3

4 Why You Should Care The web really works because it is interoperable. What s true for human-rendered data is true for machine-readable metadata. So how can web publishers make metadata interoperable? 4

5 Talk Outline What is Web Interoperability? Technology for Web Metadata Real-World Examples Principles of Interoperable Metadata 5

6 Interoperability 6

7 Technorati Yahoo Ulrich 7

8 Gada.be Technorati 8

9 Single-App Interoperability 9

10 Lessig Blog Slashdot Boing Boing 10

11 11

12 Multi-App Interoperability 12

13 Lessig Blog 13

14 a Metadata Nirvana 14

15 shutr Web Publishers Tool Builders MyCal Desktop Calendar DeskBlog All Your Blogs Users Ulrich Ursula 15

16 Meeting the Players 16

17 Web Publishers shutr 17

18 Tool Builders DeskBlog All Your Blogs MyCal Desktop Calendar 18

19 Users Ulrich Ursula 19

20 They all want to share data as easily as possible. 20

21 21

22 <h2>great New Sushi Place</h2> <p> it was amazing, I tell you. The freshest fish I've seen in years. And the toro... </p> licensed under a <a href= >CC License</a>. 22

23 shutr shutr Attribution: 23

24 24

25 DeskBlog All Your Blogs Ursula DeskBlog 25

26 Ursula License: Author: Ulrich 26

27 MyCal Desktop Calendar Ursula MyCal 27

28 Principles of Interoperable Metadata 28

29 Principle: Publisher Independence shutr Trackbacks URL commenting on Camera and Lens Info Resolution, Size, etc.. 29

30 Principle: Data Reuse <h2>great New Sushi Place</h2> <p> it was amazing, I tell you. The freshest fish I've seen in years. And the toro... </p> licensed under a <a href= >CC License</a>. 30

31 Principle: Self-Containment License: Author: Ulrich 31

32 Principle: Schema Modularity shutr DeskBlog All Your Blogs 32

33 Principle: Evolvability shutr DeskBlog All Your Blogs Equivalence 33

34 Publisher Independence Data Reuse Self-Containment Schema Modularity Schema Evolvability 34

35 Technology for Web Metadata XML RDF Microformats GRDDL RDF/A 35

36 XML shutr shutr 36

37 <head> <title>ulrich s Photos</title> <link rel= alternate href= /ulrich/photos/123.xml type= application/xml /> </head>... <img src= /ulrich/photos/123.jpg /> Beautiful Sushi. Taken on January 15th, licensed under a <a href= 2.5/ >CC License</a>. 37

38 <?xml version="1.0"?> <Photos xmlns= >... <Photo id= 123 > <title>beautiful Sushi</title> <creator>ulrich</creator> <date> </date> <license url= 2.5/ >... </Photo> </Photos> 38

39 <Photos xmlns= schema.xsd >... <Photo id= 123 > <title>sushi</title> <creator>ulrich</creator> <date> </date> <license url= licenses/by/2.5/ >... </Photo> </Photos> <Photos xmlns= schema.xsd >... <Photo id= 456 > <title>sushi 2</title> <creator>ursula</creator> <date> </date> <license url= licenses/by-nc/2.5/ >... </Photo> </Photos> 39

40 XML Report Card Publisher Independence Data Reuse Self- Containment Schema Modularity Evolvability everyone decides their own schema data is duplicated HTML and XML are separate files attributes are not reusable XSLT could be applied 40

41 RDF 41

42 RDF - A Refresher Everything is a triple: subject, property, object. Every item is either a URI or a literal (e.g. 3, Ben ) <> cc:license < <> dc:creator Ben Adida. </images/sunset.jpg> dc:title Sunset on the Beach. 42

43 <img src= /ulrich/photos/123.jpg /> Beautiful Sushi. Taken on January 15th, licensed under a <a href= >CC License</a>.  43

44 RDF Report Card Publisher Independence Data Reuse Self- Containment Schema Modularity Evolvability everyone decides their own schema data is duplicated HTML and RDF are entirely separate attributes are reusable additional fields are fine 44

45 Microformats DeskBlog 45

46 <head profile= >... <div class= hphoto > <img class= photo src= /ulrich/photos/123.jpg /> <span class= title > Beautiful Sushi. </span> Taken on <span class= date > January 15th, </span> licensed under a <a rel= license href= licenses/by/2.5/ >CC License</a>. </div> 46

47 If some usage of the web is too nascent to have converged, we can t easily codify it, so we choose to pass on the problem. Ryan King, Microformats.org 47

48 MF Report Card Publisher Independence Data Reuse Self- Containment Schema Modularity Evolvability schemas are tightly controlled HTML rendered data is reused Same file, but WHOLE file. attributes are not reusable, though some MFs are composable. may happen with GRDDL. 48

49 GRDDL XML Schema XSLT XML XSLT HTML XSLT XML RDF 49

50 <head profile=" <link rel="transformation" href="/grddl/extract-rdf.xsl" /> </head> <div class= photo > <img src= /ulrich/photos/123.jpg /> <span class= phototitle > Beautiful Sushi. </span> Taken on <span class= photodate >January 15th, 2006.</ span> licensed under a <a rel= license href= by/2.5/ >CC License</a>. </div> 50

51 GRDDL Report Card Publisher Independence Data Reuse Self- Containment Schema Modularity Evolvability everyone decides their schema, but it all depends on the transform. transform reuses data no correspondence attributes are reusable if exposed as RDF. additional fields are fine, but controlled by the transform 51

52 RDF/A Product of W3C s Task Force on RDF-in-HTML. Joint effort of HTML and SemWeb working groups. Working Draft of the RDF/A Primer available in the next 2 weeks. 52

53 <div xmlns:cc=" xmlns:dc=" about= /ulrich/photos/123.jpg > <img src= /ulrich/photos/123.jpg /> <span property="dc:title"> Beautiful Sushi. </span> taken by <span property="dc:creator"> Ulrich </span> on <span property="dc:date"> </span> licensed under a <a rel= cc:license href= licenses/by/2.5/ >CC License </a>. </div> 53

54 RDF/A Report Card Publisher Independence Data Reuse Self- Containment Schema Modularity Evolvability everyone decides their own schema similar to MF chunk of HTML is self-contained attributes are reusable same as RDF 54

55 RDF/A Elegant Degradation Fully marked up XHTML Client software interprets as much as it can understand, structuring it for the user to see. Anything the client doesn t understand is just XHTML-rendered. 55

56 <div xmlns:rss=... xmlns:cal=... xmlns:cc=... > <h2 property="rss:title">calendr Events</h2> <a rel="rss:link" href="/">calendar</a>. <p property="rss:description">calendr Events.</p> <link rel="rss:items" href="#items" /> <ul id="#items"> <li href="#item_56">lessig - Free Culture</li> <li href="#item_57">zittrain - E2E is Dead</li> </ul> </div> <div id="item_56" about="#item_56" role= cal:vevent > <h3 property="rss:title">lessig - Free Culture</h3> <p property="rss:description"> Larry will be giving his talk on Free Culture at <a rel= cal:location href=... >CSAIL</a> on <span property= cal:dtstart > :00-05</span> for about <span property= cal:duration >1</span> hour </p> </div> 56

57 <div xmlns:rss=... xmlns:cal=... xmlns:cc=... > <h2 property="rss:title">calendr Events</h2> <a rel="rss:link" href="/">calendar</a>. <p property="rss:description">calendr Events.</p> <link rel="rss:items" href="#items" /> <ul id="#items"> <li href="#item_56">lessig - Free Culture</li> <li href="#item_57">zittrain - E2E is Dead</li> </ul> available under a <a rel= cc:license href= cc.org/licenses/by/2.5/ >CC license</a>. </div> <div id="item_56" about="#item_56" role= cal:vevent > <h3 property="rss:title">lessig - Free Culture</h3> <p property="rss:description"> Larry will be giving his talk on Free Culture at <a rel= cal:location href=... >CSAIL</a> on <span property= cal:dtstart > :00-05</span> for about <span property= cal:duration >1</span> hour </p> </div> 57

58 Real-World Examples 58

59 Creative Commons Including CC in a data model. Sometimes, users control only a small part of the HTML of their page. That HTML should be entirely self-describing! Other times, publishers want to publish XML with CC data. How can a generic CC client detect that? RDF/A for CC HTML; GRDDL for CC XML. 59

60 Libraries / MIT DSpace OAI effort to standardize archive formats. the XML schema has been forked 150 times among 800 institutions. Would RDF/A prevent forking? No, of course not, but what is compatible would remain compatible. GRDDL 60

61 Publisher Independence Data Reuse Self-Containment Schema Modularity Schema Evolvability 61

62 So... Existing Datasets: GRDDL New HTML applications: RDF/A Questions? 62

O.Curé [1 ] Mashup, Microformats, RDFa and GRDDL

O.Curé [1 ] Mashup, Microformats, RDFa and GRDDL O.Curé [2 ] Mashup A mashup application combines contents or servives coming from several, possibly heteogeneous, applications. In the case of a we site,