Information from the 4th International Semantic Web Conference, Galway, 6-10 November 2005

see Links for web site and proceedings

The notes below are mostly Tony's, but there are some from Norman.

Some of the notes below are more useful than others. Certainly in some cases I (NG) was taking notes just to try to work out what was going on -- some of the talks were pretty hard-core logic. However I've noted a couple of useful papers where the idea, and the proceedings paper, seem worth following up. Tony: are there papers you'd highlight?

attended by:

• TonyLinde
• NormanGray
• ... anyone else?

Rules Tutorial (Sunday morning)

• slides (pdf), slides (pdf) for printing

• slides with TL comments added (pdf)
• slide 46: comment made that this is an area where OWL is limited

Ontology Design Patterns Tutorial (Sunday afternoon)

• slides (ppt)
  • Alan's download site
• start page for downloads

• slides with TL comments added (pdf)
• key point that ontology is just the beginning

Uncertainty Reasoning workshop (Monday morning)

• web site
• Proceedings: complete (pdf), individual (online)

Presentations

• Is it worth a hoot? Qualms about OWL for uncertainty reasoning (Fung) : p 1 of workshop proceedings
  • useful for ways of representing probabilistic info in OWL
  • gives rise to problems

• Fuzzy Semantics for SW Languages (Mauro Mazzieri) : p 12
  • addition of fuzziness

• PR-OWL: Bayesian Ontology Language for SW (da Costa) : p 23
  • star trek scenario!: need BN for each detected starship
  • situation specific BN to reason over instances
  • probably need something other than OWL

• Discovery & Uncertainty in Semantic Web Services (Martin-Recuerda) : p 34
  • discovery of web services
    • set of notions for matching
    • need relaxation for wider matching
    • but this leads to too many services
  • F-Broker extended with incidence calculus (developed by Prof Alan Bundy, Edinburgh, 1985)

• Ontology Learning and Reasoning (Haase) : p 45
  • Extraction of domain ontologies from natural language text
    • NLP
    • Machine learning
  • LOM: Learned Ontology Model
    • Modeling primitives for diff types of ontology elements
    • Confidence & Relevance annotations to capture uncertainty
    • Transformation into expressive knowledge representation (e.g. OWL)
      • Transform elements with confidence higher than given level

• Controlling Ontology Extension by Uncertainty Concepts through Cognitive Entropy (Chaves-Gonzalez) : p 56
  • preserving ontology robustness
  • use AR systems
    • OTTER
    • MAGE4

• The Fuzzy Description Logic f-SHIN (Stoilos) : p 67
  • Extend DLs with fuzzy set theory
    • Object belongs to fuzzy set to any degree betw 0 and 1
      • Tall(Tom) = 0.7
  • Fuzzy SHIN (f-SHIN)
    • Concepts formed in same way as in SHIN

• Generic Framework for DLs with Uncertainty (Pai) : p 77
  • Propose generic framework
    • Extend all components w uncertainty
      • DL
      • KB
      • reasoning
    • Lattice based approach
  • need to prove the area using reasoning
    • they're working on adding uncertainty to reasoning

• Stratified Probabilistic DL Programs (Lukasiewicz) : p 87
  • extension of previous DL programs
  • this one: extension with stratified probabilistic DL (spdl) programs

Lightning presentations

• see p 98-109 in workshop proceedings

• P98: Modeling degrees of conceptual overlap in SW ontologies
  • missed most of this one
• P100: Choice model applying the weighted utility theory
  • Transportation economists
• P102: Ontology-based analysis of experimental data
  • BioPAX (OWL): ont describing relations among bio entities
  • Language to define scoring functions
• P104: Paraconsistent reasoning for the SW
  • Classical logic vs paraconsistent logics
  • Allow 'sensible' reasoning in presence of inconsistencies
  • Cases
    • Distrib systems
    • Diff opinions
    • Coping w change
    • Dialethias (e.g. liar's paradox)
  • On SW
    • In ontology reasoning
    • In query languages
    • Trust issue
  • Qs
    • Inconsistency often comes out of context
    • Ought to try and keep inconsistency since it can contain a lot of information
• P106: Representing probabilistic relations in RDF
  • Represent both instance (A-box) & class (T-box) level
  • Provide core vocabulary
    • prob:Clause
• P108: Statistical reasoning - a foundation for SW reasoning
  • not presented

Tuesday sessions

e-Science

11:00 Paper Session Ic, Inis Mór Ballroom 3
Research / Academic Track Papers I-C: E-SCIENCE
Chair: Terry Payne

TL notes (.doc file from OneNote) : will try to update wiki later

"A Little Semantic Web Goes a Long Way in Biology"

Katy Wolstencroft, Andy Brass, Ian Horrocks, Phillip Lord, Ulrike Sattler, Robert Stevens, Daniele Turi

I think this is a USEFUL PAPER, because it's a vivid example of what a reasoner does for you, how little exotic logic you need, and how much you can get away with in terms of a slightly shaky ontology.

• Biology
  • Search for functional and structural domains
  • Use the presence of domains to classify proteins
  • Tools: BLAST and InterproScan. Tools show presence of domains, but do not themselves classify. For that, need experts.
  • Classification is the first step in understanding
  • Essential for comparative genomics
  • More data than can be handled, so lots of data remains uncharacterised
  • This project focused on the protein phosphatase family: large, important family (removes phosphate groups from molecules). Currently well classified -- needed to demonstrate utility
• Need QCR (qualified cardinality restrictions -- ie, `exactly two instances of protein X'). OK in DAML+OIL, hard in OWL, so needed workarounds. Resulted in modification to Protégé OWL, since RACER could support them already.
• Used InterproScan tool. Part of a myGrid workflow.
• Ontology classification performed as well as expert classification. Plus found one previously characterised but omitted from classifications, plus one previously uncharacterised (and evolutionarily conserved)

I talked to Katy later:

`a little ontology goes a long way': yes, they stretched OWL (wanted QCR), and stretched the instance store, but what she means is not that but the observation that they didn't use terribly sophisticated logical expressions when specifying the class relations.

Yes, GO has bits of silliness in it, but they don't matter, because: (i) it's better than what preceded it, and it's mostly used by humans, albeit indirectly, (ii) if folk find errors or garbage, they get on to the GO curators and fix it (like wikipedia), so it's continually being improved. (iii) while `the exterior of a cell is part_of the cell' is silly, biologists know what it means, and don't end up coming to wrong conclusions as a result of this stuff. (iv) It's evolutionary -- relax! (vi) Barry said that it was OK because a dodgy ontology worked adequately when you were doing statistical things; Katy agreed, but suggested that this wasn't a big deal, and reiterated that (i) it was better than what there was before, and that's the really important thing.

Yes, GO really does have tens of changes per day, both additions and changes/fixes, and it's grown from 1000 concepts four years ago to whatever it is now. It changes both because there are errors and incompletenesses fixed, and because biology itself changes.

(NormanGray)

"Provenance-based Validation of E-Science Experiments"

Sylvia C Wong, Simon Miles, Weijian Fang, Paul Groth, Luc Moreau

• Workflow validation: look at properties of the experiment. Requires reasoning over domain-specific knowledge.
  • static validation -- analyse the process before it runs
  • verify that data values satisfy constraints during execution -- interface matching
  • But workflow may change or have changed or not be accessible to the person doing the validation. Interfaces may be underspecified (bio services are often string-in-string-out).
• Provenance: causal (what caused what), factual (what actually happened, including failures, as opposed to what was planned), attributable (who asserted it)
• Each step in the workflow documents its inputs and output, causal and functional relations, and state, in an `provenance store'.
  • Ask: did I perform each service on the type of data that the service was intended to analyse. Compatible?
  • So can do type-consistency checking straightforwardly by post-hoc analysis of the provenance store. Types are in a hierarchy, so need some inferencing.

(NormanGray)

• Simon Miles' list of publications

"Seven Bottlenecks to Workflow Reuse and Repurposing"

Antoon Goderis, Carole Goble, Ulrike Sattler, Phillip Lord

• xxx

Semantic Web Services

13:30 Paper Session IIc, Inis Mór Ballroom 3
Research / Academic Track Papers II-C: SEMANTIC WEB SERVICES
Chair: Sheila McIlraith

TL notes (.doc file from OneNote) : will try to update wiki later

“Information Modeling for End to End Composition of Semantic Web Services”

Arun Kumar, Biplav Srivastava, Sumit Mittal

• xxx

“Choreography in IRS-III – Coping with Heterogeneous Interaction Patterns in Web Services”

John Domingue, Stefania Galizia, Liliana Cabral

A USEFUL PAPER, I think, not for resource discovery, but as an alternative way to approach the problem of remote applications.

John Domingue, IRS-III

• Background
  • IRS-III is intended to be a broker between a desire/intention and interactive WS. So there's an ontological separation between user and WS contexts -- two worlds.
  • Need support for discovery, composition, mediation ($1 on programming -> $5-$9 on integration, according to IBM)
  • Mediate Data, Protocol, Business process. This is about the protocol mediation.
  • WSMO: ontologies, goals, WS (description by capability=functional, and interfaces=usage) and mediators.
• Choreography
  • IRS is a broker. State is important (message-based, so state important).
  • Implemented in WSMO, so you can reimplement it
  • Comms model inspired by KADS, plus {obtain,present}-initiative
  • Can suspend choreography while awaiting further information.
  • Aimed to be easy to use by folk who aren't SW expects.
• Demo of the same booking-train-tickets, which is basically the same as in the Sunday workshop.
• Take-home
  • broker requires mediation of data/protocol/process
  • choreography -- all possible interactions with a WS
  • client choreography -- how to achieve a speciic functionality.
• Questions
  • how large have you tried? Nothing very large yet. Mostly use-cases. 10s of WS. Aiming currently to get it to work by use-case partners. Others come later.
  • Who provides the client choreography description? Didn't fully catch the answer.. Client is tied to the WS, so if you change the latter, you need to change the former -- poss, maintainance problem.

(NormanGray)

• John Domingue's page
• IRS project page

“Web Service Composition with Volatile Information”

Tsz-Chiu Au, Ugur Kuter, Dana Nau

• xxx

“Towards a Formal Verification of OWL-S Process Models”

Anupriya Ankolekar, Massimo Paolucci, Katia Sycara

• xxx

Web Services

15:30 Paper Session IIIc, Inis Mór Ballroom 3
Industrial Track Papers III-C: WEB SERVICES
Chair: Alain Leger

TL notes (.doc file from OneNote) : will try to update wiki later

“Automated Business-to-Business Integration of a Logistics Supply Chain using Semantic Web Services Technology”

Chris Preist, Javier Esplugas Cuadrado, Steve Battle, Stuart Williams, Stephan Grimm

• xxx

“Ontological Approach to Generating Personalized User Interfaces for Web Services”

Deepali Khushraj, Ora Lassila

• xxx

“An Application of Semantic Web Technologies to Situation Awareness”

Christopher Matheus, Mieczyslaw Kokar, Kenneth Baclawski, Jerzy Letkowskios

• xxx

Wednesday sessions

Agents and Distributed Architectures

10:30 Paper Session IVb, Inis Mór Ballroom 2
Research / Academic Track Papers IV-B: AGENTS AND DISTRIBUTED ARCHITECTURES
Chair: Alun Preece

TL notes (.doc file from OneNote) : will try to update wiki later

“A Strategy for Automated Meaning Negotiation in Distributed Information Retrieval”

Vadim Ermolayev, Natalya Keberle, Wolf-Ekkehard Matzke, Vladimir Vladimirov

• xxx

“Introducing autonomic behaviour in semantic web agents”

Valentina Tamma, Ian Blacoe, Ben Lithgow Smith, Michael Wooldridge

• xxx

“An ontological framework for dynamic coordination”

Valentina Tamma, Chris van Aart, Thierry Moyaux, Shamima Paurobally, Ben Lithgow Smith, Michael Wooldridge

• xxx

RDF Querying and Storage

13:00 Paper Session Va, Inis Mór Ballroom 1
Research / Academic Track Papers V-A: RDF QUERYING AND STORAGE
Chair: Bijan Parsia

TL notes (.doc file from OneNote) : will try to update wiki later

"Benchmarking Database Representations of RDF/S Stores"

Yannis Theoharis, Vassilis Christophides, Grigoris Karvounarakis

• xxx

"Containment and Minimization of RDF/S Query Patterns"

Giorgos Serfiotis, Ioanna Koffina, Vassilis Christophides, Val Tannen

• xxx

"BRAHMS: A workBench RDF store And High performance Memory System for Semantic Association Discovery"

Maciej Janik, Krzysztof Kochut

• xxx

Ontology Creation

15:00 Paper Session VIa, Inis Mór Ballroom 1
Research / Academic Track Papers VI-A: ONTOLOGY CREATION
Chair: Marta Sabou

TL notes (.doc file from OneNote) : will try to update wiki later

“RelExt: A Tool for Relation Extraction from Text in Ontology Extension”

Alexander Schutz, Paul Buitelaar

• xxx

“Graph-based inferences in a Semantic Web Server for the Cartography of Competencies in a Telecom Valley”

Fabien Gandon, Olivier Corby, Alain Giboin, Nicolas Gronnier, Cecile Guigard

• xxx

“Ontology Design Patterns for Semantic Web Content”

Aldo Gangemi

• xxx

Thursday sessions

Rules

Paper Session VIIa, Inis Mór Ballroom 1
Research / Academic Track Papers VII-A: RULES
Chair: Ian Horrocks

TL notes (.doc file from OneNote) : will try to update wiki later

"Combining RDF and OWL with Rules: Semantics, Decidability, Complexity"

Herman ter Horst

• Unifying RDF semantics with rules, focusing on decidability and complexity. Hang on -- it's going to be logical.
• RDF interpolation lemma... no, I'm lost!
• OWL has iff-semantics, RDFS only if-semantics (if c is a subclass of d in RDF, then each instance of c is an instance of d; in OWL, c is a subclass of d iff each instance of c is an instance of d). Lots of issues about complexity and decidability
• I don't think this is going to make the VO hum like a well-greased machine...

"Stable Model Theory for Extended RDF Ontologies"

Anastasia Analyti, Grigoris Antoniou, Carlos Viegas Damasio, Gerd Wagner

• Lots of detail about four-valued logic

• Aim to extend RDF to cope with negative information, derivation rules, and uncertain information
• Handle strong negation for explicit falsity, and negation-as-failure (queries absence of info)
• Stable models provide a declarative (model-theoretic) semantics for NAF
• ERDF is RDF plus strong negation, NAF, support predicates for open and closed word reasoning, derivation rules express definitions, default rules and heuristics
• Two kinds of negation:
  • Strong: -likes(Gerd, Merlot) = Gerd dislikes Merlot (that's a `not' sign)
  • NAF: ~likes(Gerd,Merlot) = Gerd does not like Merlot (that's a tilde)
  • Hmm -- these aren't really different as sentences
• Partial logic: true, false, undertedermined (truth-value gaps), overdetermined (truth-value clashes=paraconsistency). Simplest conservative extension of classical logic that allows two negations.
  • Strong negation: -u=u
  • Weak negation: ~u=t (sim, but not the same as NAF, becomes the same (?) with stable semantics)
  • guest(x) & wineForDinner(y) -> likes(x,y) vs guest(x)&wineForDinner(y)->~-likes(x,y) (no-one rejects the wine)
• and so on...

"Resolution-Based Approximate Reasoning for OWL DL"

Pascal Hitzler, Denny Vrandecic

• xxx

Versioning and Context

Paper Session VIIIb, Inis Mór Ballroom 2
Research / Academic Track Papers VIII-B: VERSIONING AND CONTEXT
Chair: Jeff Heflin

TL notes (.doc file from OneNote) : will try to update wiki later

“Reasoning with Multi-Version Ontologies: a Temporal Logic Approach”

Zhisheng Huang, Heiner Stuckenschmidt

• Describing a logic for reasoning which includes changes in the ontology, resulting from versioning.
• That means that you describe operators like Prev(phi) (meaning that phi was true in the previous version) and P(phi) (phi was true in some previous version), H(phi) (...in all previous versions)
• Eventually I realised that the goal was to reason about the ontology, not with it, and that the thing of interest to the community was the fact that a consistent logic for doing this could be devised, rather than that there was anything particular that the author could now do.

“Decentralized Case-Based Reasoning for the semantic Web”

Mathieu d'Aquin, Jean Lieber, Amedeo Napoli

• xxx

Semantic browsing of digital collections

Trevor Collins

This was a really interesting and USEFUL PAPER

• CIPHER -- Communities of Interest Promoting Heritage in European Regions. Aimed to encourage and improve, rather than replace, visits to museums
• Museum visits are `free-choice learning' -- ie, for interest rather than assessment.
• Search goal categories: navgational (find known pages) 13%; resource (find a resource/service) 25%; informational (learn something new) 62%.
• Presenting search results: category interfaces better than lists, when more than one thing is sought
• Collections: `an organised collection of objects forms a narrative that expresses relationships across the included items' (Pearce, 1995)
• Bletchley Park museum: exhibits have labels attached, text the label to a number, then go home and to www.bletchleypark.org.uk <http://www.bletchleypark.org.uk/>, then put in their mobile number to get a collection of items of interest.
• CIDOC Conceptual Reference Model, plus a story-and-narrative ontology, plus a domain ontology.
• SMS text terms turned into concepts in this ontology, retrieve classes. Easy.
• Form connections, given the list of items the user has declared themself interested in
  • eg: connect Alan Turing to Block G. Find items referred to in descriptions of both. eg: Alan Turing designed the bombe, which was used by someone working in Block G.
  • This could be useful for resource discovery!
  • Categories
  • Horizontal links are formed between categories using fact triples. Categories `has-actor alan-turing', or `has-work-location hut-6' (alan-turing was-head-of hut-6)
  • Structured output is based on the semantics
• Selection, organisation, exploration cycle. Folk can find the things they're interested in, then organise them.
• Can be used without visiting Bletchley.
  • SMS text to +44 7985 400413 example turing blockg churchill uboat polish
  • http://www.bletchleypark.org.uk/text

The really interesting thing about this, I think is the `connect Alan Turing to Block G' step, as it seems a very straightforward and powerful application of reasoning, to create a link that is implicit but unrecognised between pieces of data which are separately simply described. That seems to be to me exactly the sort of semantic resource discovery we're after!

Plenary session

TL notes (.doc file from OneNote) : will try to update wiki later

Semantic Web Challenge Award 2005

• Won by www.confoto.org <http://www.confoto.org/> -- conference photos.
• Lessons
  • Ontology reuse of shallow ontologies (SKOS, DC, ACM, FOAF, ...),
  • with complex ontologies only in specialised fields
  • little deep reasoning required to produce a nifty application -- quick, shallow, stuff is enough

RuleML Opening Remarks

• xxx

Invited Talk: Putting the Web back in Semantic Web

Sir Tim Berners Lee, Director, World Wide Web Consortium

• Slides
• Mentioned by Tim in his talk: W3C Creates Rule Interchange Format WG

ISWC Closing Remarks

• xxx

Links

ISWC 2005

• Web page
• Proceedings (downloadable in pdf form)

ISWC 2004

• Web page
• Proceedings (downloadable in pdf form)

ISWC 2003

• Web page
• Proceedings (downloadable in pdf form)

Others

• Semantic Web Science Association
• semanticweb.org
• W3C Semantic Web
• Journal of Web Semantics: Preprint Server