Multi-agent Semantic Web Systems: RDFa . . . . . Michael - - PowerPoint PPT Presentation

. . . . . .

. . . . . . .

Multi-agent Semantic Web Systems: RDFa

Michael Rovatsos

School of Informatics

21 February 2011

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 1 / 22

. . . . . .

. . .

1

Intro to Microformats . . .

2

RDFa

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 2 / 22

. . . . . .

Review

RDF(S) implements part of the vision of the SemWeb: Obtaining network effect for SemWeb requires overcoming the ’Semantic markup bottleneck’. Promising approaches:

◮ Using NLP to convert unstructured information into RDF(S) ◮ Microformats / RDFa for embedding RDF into web documents ◮ Exporting from RDBs to RDF stores. Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 3 / 22

. . . . . .

Semantic Markup Embedded in HTML, 1

Existing document web is firmly established — can we use this base to support semantic applications? General idea: embed semantic markup into existing (X)HTML tags. .

Example: hCard

. . . . . . . . <span class="tel">0131 650-1234</span> .

Example: RDFa

. . . . . . . . <span property="foaf:phone">0131 650-1234</span> Perspective 1: This is a regular web page, but hey, we can get semantics out

• f it.

Perspective 2: This is machine-readable data, but hey, humans can browse it just like the Good Old Document Web.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 4 / 22

. . . . . .

Semantic Markup Embedded in HTML, 2

Approach is best-suited to information which fits into familiar and frequently used categories, e.g., contact information, calendar events, geo-location information, product reviews, social network information. Why do we need more than existing HTML tags?

◮ In fact, has been explored: cf. “screen scraping”, “wrapper induction” ◮ But traditional HTML tags primarily oriented towards visual rendering. Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 5 / 22

. . . . . .

Semantic Markup Embedded in HTML, 3

What about HTML’s META tag? .

META in Informatics Web Page

. . . . . . . . <META NAME="DC.Creator" CONTENT="Manuel Pita"> <META NAME="DC.Creator.Address" CONTENT="m.marques-pita@ed.ac.uk"> <META NAME="DC.Publisher" CONTENT="School of Informatics, ..."> But: META information is invisible to humans Tends to fall out of synchronization with visible content Has been abused by people concerned with SEO (and therefore can actually harm your Google rank).

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 6 / 22

. . . . . .

Microformats, 3

Microformats = Semantic Web for the Impatient ? But so far, efforts to infuse the Web with meaning have gained little traction. These initiatives have been bogged down by complexity and over-ambitious goals, or have simply been too much trouble to implement at a large scale. Proponents of microformats subscribe to general idea of “web of data”, but unconvinced by RDF, OWL, etc. Microformats intended to lower “barrier to entry” — no need to learn a new language. Try to use visible data rather than invisible metadata. Try to re-use existing standards / vocabularies for data modelling; e.g., vCard, iCalendar ⇒ hCard, hCalendar.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 7 / 22

. . . . . .

RDFa

RDFa can be seen as a new way of serializing RDF. RDFa = RDF_in_attributes Integrates into XHTML, but simpler than RDF/XML. Motivation is similar to microformats, but

◮ more generic; ◮ can benefit from many existing tools for processing RDF.

Two basic cases, corresponding to whether object is literal or resource. subject predicate

• bject

literal as object about property PCDATA or content URI as object about rel href

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 8 / 22

. . . . . .

RDFa = RDF + XHTML, Case 1

element: div text: NLP with Python attribute: class title isbn:51649-9 NLP with Python dc:title element: div text: NLP with Python attribute: class title attribute: about isbn:51649-9 attribute: property dc:title

RDF triple XML node tree XML node tree + RDFa

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 9 / 22

. . . . . .

RDFa = RDF + XHTML, Case 1

subject

• bject (literal)

predicate

element: div text: NLP with Python attribute:class title attribute:about isbn:51649-9 attribute:property dc:title

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 10 / 22

. . . . . .

RDFa, Case 1

.

RDF triple

. . . . . . . . isbn:51649-9 dc:title "NLP with Python" . .

XHTML

. . . . . . . . <div class="title"> NLP with Python </div> .

XHTML with RDFa

. . . . . . . . <div class="title" about="isbn:51649-9" property="dc:title" > NLP with Python </div>

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 11 / 22

. . . . . .

RDFa = RDF + XHTML, Case 2

element: div text: Ewan Klein attribute: class author isbn:51649-9 dc:creator element: div text: Ewan Klein attribute: class author attribute: about isbn:51649-9 attribute:rel dc:creator

RDF triple XML node tree XML node tree + RDFa

• rm:ewan_klein

attribute: href

• rm:ewan_klein

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 12 / 22

. . . . . .

RDFa = RDF + XHTML, Case 2

isbn:51649-9 dc:creator

element: div text: Ewan Klein attribute: class author attribute: about isbn:51649-9 attribute:rel dc:creator

• rm:ewan_klein

attribute: href

• rm:ewan_klein

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 13 / 22

. . . . . .

RDFa, Case 2

.

RDF triple

. . . . . . . . isbn:51649-9 dc:author orm:ewan_klein . .

XHTML

. . . . . . . . <div class="author"> Ewan Klein </div> .

XHTML with RDFa

. . . . . . . . <div class="title" about="isbn:51649-9" rel="dc:creator" href="orm:ewan_klein"> Ewan Klein </div>

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 14 / 22

. . . . . .

Cafe Web Page

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 15 / 22

. . . . . .

Cafe Web Page, 1

.

File Metadata, 1

. . . . . . . . <div about="" rel="dc:creator" resource="ihp:ehk" /> Invisible in this case, so vulnerable to previous criticism. Empty value of about refers to current document by default.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 16 / 22

. . . . . .

Cafe Web Page, 2

.

File Metadata, 2

. . . . . . . . <div class="license"> All content on this site is licensed under <a rel="license" href="http://creativecommons.org/licenses/by/3.0/"> a Creative Commons License </a>. </div> RDFa slots easily into an a element;

• ther examples of SPO with URI object extrapolate from this.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 17 / 22

. . . . . .

Cafe Web Page, 3

.

Image Metadata

. . . . . . . . <img src="mmwalk.jpg" about="mmwalk.jpg"> <span rel="dc:creator" href="ihp:ehk" /> <span property="dc:subject" content="View of Middle Meadow Walk" /> <span rel="foaf:depicts" href="db:Edinburgh" /> </img> Object of the Dublin Core subject relation expressed via a literal. But no suitable PCDATA (i.e., text) daughter, so use value of content attribute instead. All spans that are subordinate to the img element share their parent’s specification for about.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 18 / 22

. . . . . .

Cafe Web Page, 4

.

Table Row, 1

. . . . . . . . <tr> <td> <a about=":aroast" typeof="gr:BusinessEntity" rel="rdfs:seeAlso" href="http://www.artisanroast.co.uk/"> Artisan Roast </a> </td> ... </tr> Uses GoodRelations (gr prefix) ontology Special syntax for rdf:type.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 19 / 22

. . . . . .

Cafe Web Page, 4

.

Table Row, 2

. . . . . . . . <tr typeof="v:Review"> <td rel="v:itemreviewed" href=":aroast"> ... </td> ... </tr> Makes the whole row map to a blank node subject of type v:Review. This is a vocabulary promoted by Google (cf. link on MASWS Wiki page for RDFa) We have to repeat the URI of Artisan Roast to provide object for v:itemreviewed.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 20 / 22

. . . . . .

Cafe Web Page, 5

.

Table Row, 2

. . . . . . . . <tr typeof="v:Review"> ... <td about=":aroast" property="v:address"> Broughton Street </td> <td property="v:rating" content="5">*****</td> ... </tr> Hierarchical structure of the table row doesn’t mesh nicely with the fact that we have two subjects:

◮ the review ◮ the business

We use content attribute to normalize the rating literal.

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 21 / 22

. . . . . .

Cafe Web Page, 6

.

Extracted RDF

. . . . . . . . <:aroast> a gr:BusinessEntity ; v:address "Broughton Street" ; rdfs:seeAlso <http://www.artisanroast.co.uk/> . [ a v:Review ; v:itemreviewed <:aroast> ; v:rating "5" ; v:summary "Best coffee ..."].

Michael Rovatsos (School of Informatics) Multi-agent Semantic Web Systems:RDFa 21 February 2011 22 / 22