ISWC 2006

From LeibnizWiki
Jump to: navigation, search

This is just to inform you about some of the interesting (and fun) things I came accross at the ISWC 2006 and OWLED 2006 conferences.

Websites: http://iswc2006.semanticweb.org and http://owl-workshop.man.ac.uk/OWLWorkshop06.html

I recommend anyone reading this to check out the Semantic Web Challenge finalists!

Note: All data about the conference, accepted papers etc, is available as RDF, and can be queried through the SPARQL Endpoint or the SPARQL Query Interface


Contents

Day 1: Terra Cognita workshop

Website: http://www.ordnancesurvey.co.uk/terracognita

This workshop focused on the many interesting issues involved with incorporating geospatial information on the Semantic Web.

Invited Talk

The invited talk was by Max Egenhofer of the Spatial Information Science & Engineering group, University of Maine, who touched upon the following issues:

Two perspectives

  • Semantics of geospatial information (i.e. explicit geospatial operations, spatial search, map-like operators)
  • Semantics with geospatial information (i.e. implicit geo operations, location as a glue between non-spatial properties)

Current situation puts too much emphasis on:

  • Languages to describe ontologies (i.e. GML-like efforts), and
  • Design of ontologies

In his opinion the community should move from how to express meaning, to how to measure meaning: geometry. He therefore introduces the notion of ontometry: the geometry of semantics, calculations over geo-ontologies. This is to consist of:

  • orders of magnitude reasoning
  • combinations of meanings

Open Issues

  1. What to specify about spatial operations
  2. Comparison of geospatial ontologies
  3. Comparing geospatial datasets
  4. Regional terminology: Brötchen, Schrippe, Semmel
  5. Living with inconsitencies
    What does 'inconsistency' mean in a geospatial context. What about updates over time..?
  6. Semantic differences and commonalities between geographic space and other spaces (and times).

Lesson learned

GIS
learn from interoperability failure in the '90s: it cannot be done quickly, and it cannot be standardized before we know what it is.

Approach

  • Explore unorthodox approaches
  • Establish theories rather than collections.

"What's so special about spatial"

Glen Hart of Ordnance Survey.

  • Geo is often a thread of connectivity in many domains, but it is hardly ever the central theme
  • About 80% of applications have a geo component, but it is mostly quite small.

He identifies a number of issues:

  • Spatial relations
    • Explicit Topology (hard to compute, enormous overhead)
    • Implicit Topology
  • Fuzzy features
    • Lack of crisp boundaries
    • Lack of defined boundaries
    • Lack of data (e.g. precise location of an accident)
  • Contextual relations
    • House 'next to' road, pub 'near' house. The context determines granularity.
  • Fuzzy concepts
    • Man-made versus natural phenomena, viewed from different contexts. Often agreement at man-in-the-street level, but specialists' definition may vary. Many phenomena have never been defined.

The good news about Semantic Web technologies:

  • Address interoperability
  • Explicit topology
  • Process of ontology creation helps understanding

The bad news about these technologies is

  • Difficult to mix semantic and spatial queries
    • Initial spatial filtering
    • Lack of explicit topology for reasoning
    • Tools not yet available (Spatial SPARQL?)
  • Geographical fuzziness challenges precise semantics
  • Scalability: triple-store representation of geospatial data is not efficient, reasoning almost impossible.

Spatial is special:

  • Threads of interoperability
  • Requires more complex indexing
  • Large quantities of implicit ontology in data
  • Lots of fuzziness in both featurs and relations


"Fire alerts on the geospatial semantic web"

Graeme mcFerren, South Africa

Uses semantic web technologies for interpreting sensor date:

  • Early warning
  • Energy management (power surges)
  • Biodiversity management
  • Assessment of climate changes

For this they use OpenGIS Consortium Sensor Web Enablement (SWE) interfaces. Currently developing a semantically enabled version of their system (AFIS 2.x)

General Remarks

  • OpenGis Consortium (OGC) compliance is very important.
  • Finding events close in time & space, important issue.
  • Modular approaches (OWL Time...)

Day 1: Semantic Web User Interaction Panel

Panelists: Nigel Shadtbolt, Tim Berners-Lee, David R. Karger, Jim Hendler, Daniel Weitzner

David Karger
"The Pathetic Fallacy of RDF" is graph visualisations.
WSRI
Web Science Research Initiative
Interesting Software
Active Futures, Haystack, MSpace, Tabulator (FOAF)
Ontology Pipes
To funnel large quantities of data (Nigel Shadbolt) -> http://www.garlik.com
UI Stuff
Fresnel, DSpace
Life Cycle
Jim Hendler made the point that the semantic web should cater for many different users, i.e. from webmaster to grandma.

Day 2: First ISWC day

Scalability: A-Box Summarization

Research by the IBM Watson Research Center

Large A-Boxes are hard to handle, however they contain a large overhead.

A-box summary
Mapping/Indexing techniques of individuals
Scalability via filtering
Inconsitencies are mostly caused by universal or maximum cardinality rectrictions. The trick is to filter other concepts out. Even more expressive filtering is possible.

Dramatic effects (!): From millions of role assertions to 1 or 2 after filtering.

OntoWiki

The OntoWiki: a wiki-like ontology editor. See http://powl.sf.net or http://3ba.se for a demo. Contrast with Semantic MediaWiki, see http://ontoworld.org.

SPARQL End-point

Also, a term that was mentioned quite frequently is a 'SPARQL End-point'. This is a standard SPARQL interface of an RDF store (e.g. Sesame), to which clients can connect.

AKTiveRank Ontology Ranking

Work done by Christopher Brewster, at the Open University KMI on ranking of ontologies (any owl file available on the web). Existing tools such as Swoogle rank ontologies in a way similar to Google. However, this is not really a measure of how appropriate an ontology is to the users' query.

AKTive Rank does a great job, my own ontology on universities and students comes in third! (See http://www.leibnizcenter.org/~hoekstra/aargh.owl)

User Interfaces

/facet

/facet is a Web 2.0 (AJAX) framework developed at CWI (part of the MultimediaN/E-Culture project) for faceted browsing of RDF triple stores. It allows for constructing queries simply by selecting relevant 'variables' from lists. It is similar to the Flamenco system (UC Berkeley) used by http://express.ebay.com and to Longwell (http://simile.mit.edu/longwell) and BrowseRDF (http://www.browserdf.org).

Faceted browsing is query construction: constructing & traversing a decision tree.

Website: http://slashfacet.semanticweb.org

Fresnel

Fresnel is a browser-independent presentation vocabulary for RDF, developed by the Semantic Web Interst Group at W3C. Fresnel's two foundational concepts are lenses and formats. Lenses define which properties of an RDF resource, or group of related resources, are displayed and how those properties are ordered. Formats determine how resources and properties are rendered and provide hooks to existing styling languages such as CSS.

Lenses & formats that work together can be combined in groups. There has been some effort to implement a jena/sesame java parser for fresnel. Fresnel enabled applications are Longwell, IsaViz and a geonames-browser.

Interesting in this light: ActiveRDF is an AJAX (Ruby) gem for accessing RDF data.

Note: We might want do investigate the use of Fresnel for MetaLex applications. This could really leverage browsing and search of MetaLex RDF repositories.

OWL 1.1 & RDF

Had a chat with Holger Knublauch about the future support of OWL 1.1 in TopBraid Composer. As TopBraid uses a triple-b ased representation (Jena) as core content model of RDF/OWL files, the adoption of OWL 1.1 is a problem since currently no mapping from OWL 1.1 to RDF exists (In particular: annotations on restrictions and QCRs). Furthermore, punning might break the semantics of RDF(S).

Note: This is not to say there will never be an RDF syntax for OWL 1.1. Also the issue wrt. RDF(S) semantics does not seem to be very problematic for us. There is a discussion thread about this on the public-owl-dev@w3.org mailinglist: http://lists.w3.org/Archives/Public/public-owl-dev/ (Okt-December 2006)

Poster Sessions

Semantic Versioning Manager
SVM is a Protege plugin for versioning of OWL ontologies developed by Siegfried Handschuh (who is very interested in a good case study, e.g. in Law!). More information on this tool at http://ontoware.org/projects/semversion/
Emerges
Spatial data integration with semantic web services (one of the finalists of the Semantic Web Challenge). Part of the DIP project. More information and a demo at http://irs-test.open.ac.uk/sgis-dev/
DBin
An annotation tool for semantic web communities. Allows for easy creation of discussion groups, in Eclipse as well as on the web (publishing). Also one of the challenge finalists. See http://www.dbin.org
Foafing-The-Music
Integrate multiple feeds about music into one interface. Combines information about your musical interests (Blogger FOAF profile, Last.fm) new releases (Amazon.com, iTunes) and information about local concerts. See http://foafing-the-music.iua.upf.edu
DIP Project
Semantic web services for eGovernment in Essex county. Tutorials available online. See http://dip.semanticweb.org/
Semantic MediaWiki
Extension on top of MediaWiki that addes typed wiki-links (i.e. defining properties). Categories are classes, pages are instances. Export and import to RDF. See http://ontoworld.org/

Day 3: Second ISWC day

Invited Lecture

Invited lecture by Jane Fountain, from NCDG, a political scientist.

Interoperability is not only Semantic Web, but also jurisdictional, legal and behavioral. It concerns power. Interoperability is necessary, but not sufficient for political voice.

eGovernment is vendor-driven, often many eCommerce tools are used for eGovernment applications. Governments are knowledge organisations, but are poor at providing information about their decision making processes.

Max Weber - Bureaucracy
Jurisdiction
clear scope of authority
Hierarchy
decision making rights
Rules & Procedures
equitable
Files
Precedents, documentation
Neutrality
no favorites
Replaces particularism and delimits authority.

Challenges for Public Servants

  • Vertical institutional structures
  • Perverse incentives (government managers work on budgets: budget must be kept high)
  • Misuse of capital/labor substitution
  • Outsourcing vs. integration/reform
  • Customer service strategies

Competing Logics

  • Democracy, equality
  • Security, privacy
  • System feasability, interoperability, adaptability, standardization
  • Feasibility (administrative & political)
  • Agency autonomy & flexibility
  • Economics (butgeting, resources)

Papers

Can OWL & Logic Programming live Happily ever after?

By Boris Motik et al., University of Manchester

An attempt to integrate DL & LP into one language. Current languages have some missing features:

  • OWL can only express tree-like axioms
  • No possibility to express polyadic predicates: flight(from,to,airline)
  • Cannot express exceptions, e.g. Dextrocardiacs carry their heart on the right side of their body, there is practical need to be able to express defaults
  • Closed worlds (OWL vs. LP)
    • Partial solution: "close predicate", express that "these are the only facts about flights" (complete enumeration)
    • Closed worlds vs. closed domain reasoning (e.g. everybody is either peter or paul)
  • Constraints, e.g. each person must have an SSN
  • Need to be able to reason about our own knowledge (i.e. the knowledge that exists in the KB)

First order rule formalisms:

  • Many exist (SWRL etc)
  • Semantics is standard FOL
  • Easily undecidable
    • Decidability achieved by syntactic restrictions
    • e.g. DL safety -> apply rules only to known individuals
  • Easy to extend with polyadic predicates
    • no nonmonotinic features

Auto-epistemic knowledge

  • Exceptions & constraints need introspection -> introduce the K (knowledge) operator.
  • Exceptions, when used in queries does not change semantics, an provides an algebra-like QL.

The future...: OWL + DL Safe Rules + MKNF

Note: Read the paper! http://iswc2006.semanticweb.org/items/Motik2006bh.pdf


A Model Driven Approach for Building OWL DL and OWL Full Ontologies

By Saartje Brockmans et al., AIFB Karlsruhe

  • Define a UML meta-models for OWL, RDF etc.
  • Model-driven ontology development (ODM)
  • Commonlogic module, TopicMaps module
  • UML2 profiles (for UML users) -> extension to UML (stereotypes, UML constraints)

See the ODM Specification

Note: Read the paper! http://www.aifb.uni-karlsruhe.de/WBS/sbr/publications/omgodm.pdf


Day 4: Third ISWC day

The invited talk was by Rudi Studer of the AIFB Karlsruhe. He gave an interesting overview of current research efforts and uncharted territory on the Semantic Web.

See his Powerpoint slides or a PDF abstract.

Papers

SADIe: Semantic Annotation for Accessibility

By Sean Bechhofer, University of Manchester

Project to improve the readability of web-pages for the visually impaired. Screenreaders read only text (from the HTML source), but the focus of web designers is on presentation. Because it is unlikely that web-designers will change their ways, and pages are often generated dynamically rather than static, develop a proof-of-concept that annotates the CSS.

An ontology provides knowledge about tags used in CSS... the pages are fed through a proxy which interprets the CSS and alters it to improve presentation for screen readers

Prototype transformations:

  • Defluff
  • Reordering
  • Tool menu

The system can be tested online at http://rpc238.cs.man.ac.uk, more info on http://sadie.cs.manchester.co.uk

Note: Actually rather an interesting idea to do CSS annotation rather than the pages themselves. Maybe MetaLex can benefit in some way?

GINO: Guided Input Natural Language Ontology Editor

By Abraham Bernstein and Esther Kaufmann, Universität Zürich

Uses semi-structured natural language to construct inserts (i.e. class definitions), but also for querying.

http://www.ifi.unizh.ch/ddis/research/semweb/talking-to-the-semantic-web

Ontology Visualization using 'Cropcircles'

By David Wang and Bijan Parsia.

The Cropcircles are a way to visualise an ontology tree using venn-like diagrams.

Some of the features:

  • Is in fact a tree-viewer
  • Support for multiple-inheritance
  • Told vs. inferred view
  • Trees in graph nodes
  • Partitioning of ontologies using e-connections
  • Nesting of circles, size ratio
  • Layout is topologically sensitive (i.e. sorted by size)

Cropcircles can be seen in action in SWOOP, select Fly the Mothership from the Advanced menu.

Note: Other visualisation tools are OWLViz, TreeMap, TGViz, OntoViz, Ontotrack. The last one uses the SpaceTree library, which is actually quite nice (beautiful, uses thumbnails to show size & attributes of subtree)


Day 5: First OWLED day

Matthew Horridge gave a presentation of the latest version of the Manchester OWL Syntax. Unfortunately there's no OWL 1.1 support yet in this language. But for those of us who like to type rather than click, the syntax can be very handy.

DIG 2.0

By Anni-Yasmin Turhan et al.

DIG 2.0 will have the following features:

  • Abstract specification in UML
  • Concrete bindings (e.g. XML, WSML)
  • Simple protocol
  • Support for several DLs
  • Contain a Core DIG interface + extensions

Possible extensions are:

  • Retrieve told information
  • Retraction (will be part of the core)
  • Collection of new inferences (NSI, least common subsumer etc.)
  • Query interface (conjunctive queries)

There will be a DIG 2.0 reference middleware which realises the new functionality.

Problems with OWL

By Boris Motik and Ian Horrocks

Problems with the current version of OWL are mainly on the level of expressivity limitations:

  • QCR's
  • Relational Expressivity
  • Datatype Expressivity
  • Metamodelling (although OWL Full has some support for this)

Other problems, when engineering:

  • OWL Lite is SHIF(D), but the syntax makes it very hard to express all SHIF(D) axioms
  • The combination of two OWL Full ontologies might turn into OWL Lite!

Note: Read the paper! See http://owl-workshop.man.ac.uk/acceptedLong/submission_13.pdf

GLOO

Graphical query language for OWL ontologies...

What OWL has done for geography and why we don’t need it to map read

By Catherine Dolbear et al.

One of the questions from the audience was: if you want to do topological reasoning, why don't you just couple your ontology with some spatial logic?

Note: Real question is whether the abundance of epistemological concepts on the web, or in commonsense, should prompt added expressiveness to OWL or maybe just easy extensibility. It's rather easy to say to (actual) users to adopt some obscure non-standard logic

Various

Issues raised by Alan Ruttenberg, when dealing with large quantities of data, there is need for:

  • Integrity constraints
  • Unique Names

Issues raised by Deborah McGuinness:

  • Better numerical support (i.e. modeling of prototypical values)
  • Change management
  • Collaboration support

Putting OWL in order

By Nick Drummond et al.

There is a real need to express

  • Time related events
  • Conceptually linear structures
  • e.g. protein structures

Classify proteins by recognising parts of sequences that occur in other sequences. OWL is useful for pattern-matching, reasoners are quite fast.

See http://www.co-ode.org/ontologies/lists

What about complicated structures? trees, tables...

Various

  • There was a show of hands about future support for constraints (integrity). All were in favor!
  • Protege now supports OWL 1.1 (in version 4.0a).
  • The D2R project has published the entire DBLP as RDF(S)
  • Some discussion about the unfortunate semantics of owl:imports statements (why unfortunate?). New modular semantics provides a localised interpretation of imports statements (based on e.g. e-Connections). This allows for reasoning only over the overlapping parts of two ontologies. For this declaration of class definitions in the ontologies is absolutely necessary.
  • Prova is a typed hybrid description logic programming language. http://www.prova.ws
  • Some discussion on tractable fragments, and achieving tractability through modularization.
  • Precomputing inferences, storing them in a DB to achieve performance gain.
  • A show of hands about expressiveness vs. tractability resulted in a tie: both equally important to the community.
  • Request by reasoner developers (Parsia, Horrocks) to come up with cool ontologies that either crash reasoners or contain tons of data.

About OWL 1.1

Features of OWL 1.1 are:

  • Qualified Cardinality Restrictions (QCR)
  • Property inclusion axioms
  • Reflexive, irreflexive and anti-symmetric properties
  • Disjoint object properties
  • N-ary datatype predicates (e.g. width and height should be equal)
  • User-defined datatypes
  • Possible new features will be:
    • Inverse-functional datatype peoperties
    • Relations between values for different objects (e.g. people older than their boss)
    • Aggregation functions (e.g. sum of duration of subprocesses = duration of process)
  • Declarations and structural consistencites
  • Annotations
  • Disjoint union, disjoint classes
  • Punning (class, property & individual are not disjoint)
  • OWL 1.1 is (structurally) specified using UML diagrams
  • OWL 1.1 RDF/XML is backwards compatible (not completely finished yet).

The audience mentioned need for:

  • Universal role (which is only a technical addition)
  • Inverse functional datatype properties (simple constraint checking)

It was voted to add these to a future version... (but probably not for OWL 1.1)

Day 6: Second OWLED day

Reasoning Support for Ontology Design

By Uli Sattler, University of Manchester

Making a changes in an ontology can result in the time-bomb effect. After a long time editing, adding a single restriction can result in loads and loads of unsatisfyable classes.

Propose to use some form of abduction to do diagnosis and debugging: Add explanation to theory which can explain an observation. This requires the introduction of good criteria for selecting explanations.

Brain imaging data analysis

By Yolanda Gil, see http://vtcpc.isi.edu/provenance --> search for 'wings'

Generate workflows on the basis of an workflow templates defined in an ontology. They do workflow checking by using AI planning techniques, i.e. they do not use an OWL reasoner for validation & verification of workflows.

Belief Revision in SWOOP

By Bijan Parsia, University of Manchester

SWOOP supports Belief Base Revision: look at the explicit syntax instead of the models of the theory. SWOOP does some of this... it is certainly not correct, but adopts a minimal change-rule to advise users about a restriction which might have caused the ontology to be unsatisfiable.

W3C Working Group & other decisions

It was decided that the process to set-up a W3C working group for OWL 1.1 (or some other name) should be started. This group will begin its work in the first half of 2007, and finish somewhere in 2008.

OWLED 2007
Will be held 2 days at ESWC 2007 (6-7 june) in Innsbruck, just before DL2007 (8-7 june) in Bolzano.
W3 Meeting
Will be held just before ESWC

Some of the things we decided would be really nice to have:

  • SPARQL DL Constructive queries
  • Constraints
  • Human friendly syntax (part of OWL 1.1 ?)

Requests for:

  • List of issues
  • List of things that can use our effort (SWEO?)

Day 7: Back home

Yay! Took the limo to the airport :)

Retrieved from "http://wiki.leibnizcenter.org/index.php/ISWC_2006"
Personal tools