“Taxonomy information” from Wikispecies; use XSLT to extract RDF from XHTML
Use SPARQL as an integration tool (see also demo online)
However…
Things do not always fit into this model nicely
the initial database may include “weights” for certain relations
ontologies may not cleanly separate or bind terms, there are “shades”
etc
A specific case
A specific case (cont.)
A Semantic Web application would:
encode such flowchart (“clinical pathway”) in ontologies and/or rules
combine it with data coming from other sources (drug information, hospital administration, etc…)
provide a comprehensive tool to help doctors
(Go to the HCLS Workshop tomorrow if you are interested…)
What are the proper conclusions?
Obtain chest X-ray, especially if patient has two or more of these signs:
Temp > 37.8ºC
Pulse > 100
Decreased breath sounds
Respiratory rate > 20
But, surely:
Temp > 37.5ºC is in the “danger zone” already
symptoms carry different weights in decision making
etc
Similar scenarii arise in…
Biochemical, biological research
Oil exploration
Exchanging/modelling spam control rules
Legal profession
Media content rating rules
Multilingual setting
Even the ground data may bear some weights/uncertainties:
E.g., geotagging via Google, if you are in a country not “covered” by precise cartography data
An aside…
Studies have shown that traditions in East Asian cultures (China, Korea, Japan,…) are very different
not based on the rational Greek heritage
the “shades” and the overall picture is taken more seriously (think of traditional Chinese medicine)
yes-or-no logic is less natural than for others
Ie, modelling, eg, Chinese medicinal approaches on the Semantic Web might be more difficult…
Things can be hacked, of course…
One can imagine complex modelling, e.g., (in the pneumonia example)
separate classes with various temperature intervals and with an extra property for a weight
class axioms using the various combinations of these
etc
Combine a traditional reasoning on some part of the knowledge base, and a statistical reasoning on other parts
SPARQL queries may contain FILTER-s evaluating numerical values
But all these lead easily to the equivalent of a spagetti code…
Instead…
A clean (reasoning) model may make great sense
and that is where workshops like this are important
However: it should be part of the Semantic Web landscape!
By the way…
Are we talking about probabilistic reasoning, or…
… reasoning based on fuzzy logic, or…
…both?
It seems that these two approaches (or the communities?) are fairly independent of one another…
Where are we now (reminder)?
Solid specification suite since 2004: well defined (formal) semantics for RDF, RDFS, OWL, clear RDF/XML syntax
Lots of tools are available. Are listed on W3C’s wiki:
RDF programming environment for 14+ languages, including C, C++, Python, Java,
Javascript, Ruby, PHP,… (no Cobol or Ada yet !)
13+ Triple Stores, ie, database systems to store (sometimes huge!) datasets
etc
Some of the tools are Open Source, some are not; some are very mature, some
are not :
it is the usual picture of software tools, nothing special any more!
Anybody can start developing RDF-based applications today
Where are we now? (Cont.)
Separate layers have beed defined for OWL, balancing expressibility vs. implementability (OWL-Lite, OWL-DL, OWL-Full)
SPARQL is coming to the fore, with lots of implementations already
it will play an important role in integrating/federating RDF data
Lots of developments are going on
“Scraping” RDF data from various sources: images (XMP), XHTML/XML sources (e.g., GRDDL, RDFa)
Building SQL ⇋ RDF “bridges” to export data to RDF
Developing rule interchange formats (RIF)
etc
Revisions of RDF and OWL?
Such specifications have their own life
Missing features come up, errors show up
There will probably be a next version at some point
Revision of the RDF model?
Some restrictions in RDF may be unnecessary (bNodes as predicates, literals as subject, …)
Issue of “named graph”: possibility to give a URI to a set of triplets and make statements on those
Alternative serializations (XML or otherwise)?
Add a time tag to statements?
Internationalization issues with literals (how do I set “bidi” writing?)
These are just ideas floating around…
“Light” ontologies
For some applications RDFS is not enough, but even OWL Lite is too much
There may be a need for a “light” version of OWL, just a few extra possibilities v.a.v. RDFS
There are a number of proposals, papers, prototypes around: RDFS++, OWL Feather, pD*, …
pD*, for example, has property characterization (symmetric, transitive, inverse),
class and property equivalence, and property restrictions with some or all values
This might consolidate in the coming years
Consequences for uncertainty reasoning work
The Semantic Web is more than just RDF and OWL
Any development should be part of this overall picture!
this means:
think of querying, not only reasoning (à la SPARQL)
relying only on OWL-DL or higher may not be a fully satisfactory
think whether the core RDF semantics should be changed (and how and when…)
think of how to generate core data with embedded uncertainty
looking at SQL ⇋ RDF issues may become necessary
consider the RIF (Phase II) work, it may be a good forum to look at some of the issues
Thank you for your attention!
These slides are publicly available on:
http://www.w3.org/2006/Talks/1105-Athens-IH/
in XHTML and PDF formats; the XHTML version has active links that you can follow
Uncertainty and the Semantic Web
!)