Global Application Integration with the Semantic Web

...
<foaf:Person>
   <foaf:name>Takashi OTA</foaf:name>
   <foaf:name xml:lang="ja">太田尚志</foaf:name>
   <foaf:title>Mr</foaf:title>
   <foaf:firstName xml:lang="ja">尚志</foaf...
   <foaf:firstName>Takashi</foaf...
   <foaf:surname xml:lang="ja">太田</foaf:s...
   <foaf:surname>OTA</foaf:surname>
   <foaf:nick>takot, Takashi</foaf:nick>
   <foaf:nick xml:lang="ja">たこちー</foaf:n...
   <foaf:knows>
      <foaf:Person>
         <foaf:name>SHIMIZU Noritada</foaf...
         <foaf:firstname>Noritada</foaf:fi...
         <foaf:firstname xml:lang="ja">智公...
         <foaf:surname>SHIMIZU</foaf:surname>
         <foaf:surname xml:lang="ja">清水<...
      </foaf:Person>
   </foaf:knows>
</foaf:Person>
...

W3C specifications define the RDF language as well as ontologies defining terms for class and property hierarchy and for description logic. This provides a globally grounded data fabric through which applications can interact..

Eric Prud'hommeaux
12 March, 2004
Keio University, Japan

$Revision: 1.15 $ $Date: 2004/03/20 00:10:50 $

Available at http://www.w3.org/2004/Talks/0312-rdf-status/.

These slide contain many pointers -- intended to be used as a resource after the talk is complete.

RDF and other W3C Technology

Data in RDF
Presentation in SVG, XSLT, CSS

Publication History

1997-1998 Pre-RDF ground work ...
Feb 1999 original RDF Model and Syntax Recommendation
March 1999 RDF Schema Proposed Recommendation
1999-2003 RDF infrastructure and protocol ...
2000-2004 RDF vocabularies ...
Feb 2004 suite of RDF Recommendations (syntax, concepts, semantics, primer, test cases)
Feb 2004 Rdf Schema Recommendation
Feb 2004 suite of OWL Recommendations (overview, language Guide, reference, semantics, test cases)

replaced by later publication
W3C Recommendation

Pre-RDF Publications

Oct 1997 W3C Data Formats
July 1998 Document Content Description for XML

W3C Note

Vocabulary Publications

March 2000 PICS
April 2001 Schema for Infoset
Jan 2002 Schema for P3P
Dec 2002 Evaluation and Report Language
Jan 2004 CC/PP

W3C Recommendation
W3C Note
W3C Working Draft

Infrastructure and Protocol Publications

Jun 1999 Web Architecture: Describing and Exchanging Data
Oct 1999 Cambridge Communiqué
Aug 2000 Accessibility Features of SVG (embedding RDF in SVG)
Sep 2000 Harvesting RDF Statements from XLinks
June 2003 OWL XML Presentation Syntax

W3C Note

RDF Applications

Some example RDF applications:

validation: W3C RDF Validator
ontology: rdfs, OWL, log:, Semantic Web Web Services
query: RDQL, algae, squish, ... (see survey)
logic (rules): cwm, RACER, FaCT (when use with a translator)
data servers: Jena/Joseki, TAP (semantic search), 3store
filtering/aggregation: NewsMonster, PICS
privacy and digital rights: P3P, Creative Commons
and more...: FOAF (and FOAFCORP), Annotea, RSS 1.0, wordnet, MusicBrainz, Sherpa, RDFical (ICalendar)...

Graphical Interfaces

The graph nature of RDF encourages visualization

Graphical Interfaces 2

RDF Classes are useful for classification interfaces.

Semantic Web Web Services

The Web Services Activity specifies a set of protocols for describing web services invocation.

service discovery makes web services useful.
composability (ability to re-combine services) makes web services even more useful.
web service function description similar to RDF ontology description.
some specifications (eg OWL-S) and implementations
- Web Service Composer
- MusicBrainz

Relationship to XML Activity

...
<foaf:Person>
   <foaf:name>Takashi OTA</foaf:name>
   <foaf:name xml:lang="ja">太田尚志</foaf:name>
   <foaf:nick>takot, Takashi</foaf:nick>
   <foaf:nick xml:lang="ja">たこちー</foaf:nick>
</foaf:Person>
...

RDF rests on XML technologies for one serialization. XML provides:

namespaces
internationalization (examples in this talk are in UTF-8)
precedent to rely on a vast body of web infrastructure

Namespaces

Most serializations import utility of namespaces.
cwm extends logic layer to use namespaces (log:racine).

The RDF data model is fundamentally different from XML.

XML uses elements and attributes
RDF uses nodes and relations.
XML Schema talks about data types.
- simple types: int, string, boolean, date, time...
- complex types: Address Record, Person Record
RDF Schema provides inferences.
if X is the father of Y, X is the parent of Y.

More on data models

The difference in the data models is very important.

Looking at some colloquial (intuitive) XML...

Convenient XML

We pretend XML is simple:

<Person href="ericP"
        first="Eric"
        last="Prud'hommeaux">
  <mbox href="mailto:eric+talk@w3.org"/>
</Person>

Self-describing XML

Globally scoped XML data requires more precision:

<?xml version="1.0"?>
<foaf:Person soapenc:href="http://www.w3.org/People/all#eric"
             foaf:first="Eric"
             foaf:surname="Prud'hommeaux"
             xmlns:soapenc="http://schemas.xmlsoap.org/soap/encoding/"
             xmlns:foaf="http://xmlns.com/foaf/0.1/">
  <foaf:mbox soapenc:href="mailto:eric+talk@w3.org"/>
</foaf:Person>

XML with Arbitrary Term Re-use

Re-using defined terms from other schemas requires even more precision:

<?xml version="1.0"?>
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
<foaf:Person rdf:about="http://www.w3.org/People/all#eric"
             foaf:first="Eric"
             foaf:surname="Prud'hommeaux"
             xmlns:foaf="http://xmlns.com/foaf/0.1/"
             xmlns:pim="http://www.w3.org/2000/10/swap/pim/email#">
  <pim:mbox rdf:resource="mailto:eric+talk@w3.org"/>
</foaf:Person>
</rdf:RDF>

XML Data Model

The XML Infoset view of the Person record:

RDF Data Model

The RDF view of the Person record:

<?xml version="1.0"?>
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/...
<foaf:Person rdf:about="http://www.w3.org/P...
             foaf:first="Eric"
             foaf:surname="Prud'hommeaux"
             xmlns:foaf="http://xmlns.com/f...
             xmlns:pim="http://www.w3.org/2...
  <pim:mbox rdf:resource="mailto:eric+talk@...
</foaf:Person>
</rdf:RDF>

Relationship to (XML) Web Services

XML Protocol gave Web Services a common packing system.

can be used for RPC

WSDL provides...

function prototypes (syntax)
limited message patterns (one type of semantics)

Choregraphy

Choregraphy defining more complex patterns.
Actual semantic description is needed for next steps:
- describing a service in a machine useful way,
- using those descriptions for service matching,
- leverage off subClasses/subProperties to find best fit service and preconditions.

RDF is a Knowledge Representation System

Why do we need another?

many KR systems have been explored
ontology management was major bottleneck or source of chaos
- term invention/definition tension between centralization and underspecification
URI's leverage existing social system to create unambiguous terms
URI's provide dereferencing system for acquiring latent semantics
coincides with industry interest in process process integration
coincides with world enthusiasm for the world wide web

W3C Standardizes RDF, RDF Schema, and OWL

What do these standards give us?

common model for data
expectation of partial understanding
meta data expressed in same model

Erasure/Partial Understanding/Monotonicity

...
<foaf:Person>
   <foaf:name>Eric Prud'hommeaux...
   <foaf:name xml:lang="ja">e ri...
   <foaf:title>Professor</foaf:t...
   ...
   <foaf:made>
      <rss:channel rdf:about="h...
	 <rss:title></rss:title>
	 <rdfs:seeAlso>
	 </rdfs:seeAlso>
      </rss:channel>
   </foaf:made>
</foaf:Person>
...

RDF mandates partial understanding
- partial understanding always ok
- great opportunity for re-use of tools/data:
  dc predicates permeate many other protocols, eg foaf
  dc tools can get some use out of those protocols
This example data is available to an RSS processor.
- The whole document is RDF so parsing context is known to any app.
- Other data does not change the meaning of the RSS.

Counter Example

<?xml version="1.0"?>
<vc:vCard xmlns="http://www.jabber....
   <vc:FN="Eric Prud'hommeaux"/>
   <vc:N>
      <vc:FAMILY="Prud'hommeaux"/>
      <vc:GIVEN="Prud'hommeaux"/>
      <vc:MIDDLE/>
   </vc:N>
   <rss:channel rdf:about="http://e...
      <rss:title></rss:title>
      <rdfs:seeAlso>
      </rdfs:seeAlso>
   </rss:channel>
   <EMAIL><USERID>mailto:eric+talk@...
</vc:vCard>

By the rules of XML, an RSS application may NOT interpret the RDF in the example.

The semantics of rss:channel are defined by vc:vCard root element.
The DTD or schema for vCard will not have a place for the rss:channel element
OR, it will not define the relationship to that element.
A program or XSLT stylesheet or XQuery trying to use the RSS data would have to know about the specific combination of vCard and RSS.

Non-monotonic Counter Example

<?xml version="1.0"?>
<SOAP-ENV:Envelope
 xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
   <SOAP-ENV:Body>
      <vc:vCard
       SOAP-ENV:encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
       xmlns:vc="http://www.jabber.org/jeps/jep-0054.html">
         <vc:FN="Eric Prud'hommeaux"/>
         <vc:N>
            <vc:FAMILY="Prud'hommeaux"/>
            <vc:GIVEN="Prud'hommeaux"/>
            <vc:MIDDLE/>
         </vc:N>
         <rss:channel rdf:about="http://example.org/rssChannel">
            <rss:title></rss:title>
            <rdfs:seeAlso>
            </rdfs:seeAlso>
         </rss:channel>
         <EMAIL><USERID>mailto:eric+talk@w3.org</USERID></EMAIL>
      </vc:vCard>
   </SOAP-ENV:Body>
</SOAP-ENV:Envelope>

By the rules (conventions) of SOAP encoding, the RSS element MAY be inserted into the example.

The semantics of rss:channel are defined by soap-enc: root element.
An RSS application will NOT know if the elements preceeding it or following it change the interpretation of the rss:channel element.
A program or XSLT stylesheet or XQuery trying to use the RSS data would STILL have to know about the specific combination of vCard and RSS.

Application Integration

The previous examples show how RDF promotes publication of interoperable data. This promotes application integration:

Applications like RSS monitors are defined by the data they consume.
The RSS application may know nothing about anything in the foaf namespace.
Enterprise apps can re-use apps like foaf for medical and human resorces and department organizations...
Large enterprise apps will be able to trivially share the portion of data that is relevent to both.

Single model for "meta" data

meta data expressed in same model
- same tools used for parsing, editing and storage
- inferencing fits with partial understanding of rules
  - (you don't need to know everyone's inferences to make your own)
  - heavy jobs need heavy inferencing, other jobs need no inferencing

Recent Work

Two charters are under consideration
(member review ended 9 Feb)

RDF Best Practices Working Group
- kickoff meeting at the W3C Technical Plenary
- ontology publishing guidelines, FAQs, respositories, reference.
RDF Data Access Working Group
- query plus some protocol.
- should lay groundwork for rules standard.

W3C Technical Plenary

The RDF Interest Group met March 1-2. topics included:

Data Access Working Group
Best Practices Working Group

The BPWG met March 4-5.

The HTML WG discussed RDF in HTML.

The Device Independence WG discussed RDF query and annotations.

RDF Query

Benifits of standardization:

will provide applications with uniform access to RDF data.
will enable standard query as a (web) service.

Challenges:

expressivity: how rich should the language be?
- logical operators (not, or)
- functions on literals (<, >, substr)
- aggregates (min, max, count())
relationship to (reuse of) XQuery.

RDF Rules

RDFS and OWL have semantics that must be interpreted by a programmer.

A standard expression of rules allows the machine to assume that step.

Inventor of the term defines the rules.
User of the term imports the rules and benefits from their inferences.

Research Issues

Such a large arena, hard to select.

Here are a few recurring issues:

Context, Attribution, Provenance...
interaction with conventional databases
meaning of XML complex types
representation of time in RDF graphs
Relational database experience shows this is seldom solved in a general approach.
privacy

Context, Attribution, Provenance...

Most information management systems imply some "paper trail", information about where the information comes from.

Databases, RDF included, tend to push this information in the application layer.

There is no current standard way to describe the relationship between an assertion and the document where that assertion was found. cwm uses logInclude. Most DBs keep the information "out of the model".

Interaction with Conventional Databases

efficiency -- flexibility of RDF with the efficiency of relational DBs:

store data efficiently in relational databases
give semantic web access to data in relational databases
This provides access to much of the data now used within organizations.
use RDF to add flexibility to relational data.

Database manufacturers can provide this view of data:

Oracle Network Data Model -- store and query RDF graphs.

Conclusions:

Knowledge Representation is a well-explored field.
W3C provides expertise in scalable web technology.
W3C provides a forum for industry consensus.
RDF promotes balance between chaotic freedom and standardized interoperability