Workshop on Frameworks for Semantics in Web Services (Day 1)

9 Jun 2005

Steve Battle & David Martin


Use cases session

David Martin: one presentation before lunch and more after

"Managing enterprise applications as dynamic resources in corporate semantic webs" - F. Gandon, INRIA

Fabien Gandon: slide: ACACIA in short
...slide: Corporate Web & Intranet
... slide: Corporate semantic Web
... slide: Corporate semantic Web [2]
... slide: CORESE semantic search engine
... slide: Typical query on a memory
... using SPARQL
... slide: Discover paths between resources
... slide: Approximate search
... slide: Production rules
... slide: Architecture
... slide: HCI and portal generation
... slide: Corporate distributed knowledge
... slide: DAI & DFK

<TerryP> ... use of contract net to look for services (e.g. based on an auction/bidding model)

... slide: [ unnamed, with a screen shot ]
... slide: Corporate web services
... slide: Web services and Enterprise application
... slide: Memories with a broaden scope
... slide: Corporate semantic web services
... slide: Discover and invoke a service

[ this is a demo ]

... slide: Service description
... slide: Input description & extension
... slide: Composing with memory
... slide: Corporate semantic web applications
... slide: Ongoing...
... talking about composing services
... not convinced about automatic composition

<carine> I like the idea of saving existing composition, but that might lead to security/privacy problems... and service versioning is probably essential too

<TerryP> There is a body of work on Case Based Reasoning that does this - save plans (i.e. compositions) for future adaptation/reuse

... slide: My one cent
... starting with RDFS and not OWL was the right thing
... full integration has many facets: description, but also things like SOAP & SPARQL, SW rules, SWS as WS sources

Amit Sheth: what are the difficulties in doing the mapping between services developed by different entities?

Fabien: I'm not taking the POV of the extranet

Amit: would it be considerably more difficult?

Fabien: I don't know

Matthias Flügge: [ missed question ]

Fabien: the memory is built by the corporate employees
... this is where the commitment takes part
... if someone doesn't agree with the ontology used, than the alignment happens in the UI

Matthias: I'm not sure whether this will scale

Fabien: there's a difference between what you want to formalize and what you don't

Katia Sycara: have you done any evaluation for more IR-based document retrieval?

Fabien: we've done tests on document and skill retrieval
... we had good feedback on @@@ and precision
... however, the UI wasn't user-friendly enough
... this is the gap we currently have with IR

LUNCH BREAK; back at 1.30pm

Semantic evolution of Geospatial Web services (Todd Pehle)

Todd Pehle: Slide 1 Introduction

... Slide 2: OGC (Open Geospatial Consortium)
... Opengeospatial Web Services
... Slide 3: GSW (Geospatial Semantic Web)
... Slide 4: geospatial part/web part/semantic part
... Slide 5: geospatial reasoning
... Slide 6: what the web brings to geospatial problems
... Slide 7: Interoperability
... Slide 8: Goals
... try different ontologies
... Slide 9: Example of geo query

<mnot> wonders if http://www.openstreetmap.org/ is relevant to this talk; would be nice to see open+semantic

... Slide 10: sequence
... Slide 11: GSW Ontology components
... Slide 12: experiment
... Slide 13: Example query stack
... Slide 14: Query plan in more details
... Slide 15; (cont)
... Slide 16: technologies stack
... Slide 17: questions on OWL-S
... what if the content is dynamically changing?
... we need some form of self-description/operation
... Future of objects that are moving...

Ken Laskey (MITRE): there is a 2-level experimentation. what kind of tools?

Todd Pehle: if you query several services you can get reasoning about different data
... we kicked off the experiment 1 month ago

Ken Laskey: I try to come up with examples that would not have the power otherwise

Todd Pehle: there is more value in querying more data. we just want to do one Use case and see what pitfalls there are

Amit Sheth: spatial data is very rich, how do I select a knowledge base?
... you build a special ontology, e.g. that could have the notion of "nearby"
... is this possible to develop some sort of standard in the domain of geo..?

Application of Standard Semantic Web Services and Workflow Technologies in the SIMDAT Pharma Grid - Changtao Qu

Changtao Qu: Introduction...
... SIMDAT project in EC IST FP6
... 4 main industrial sectors: Automotive, Pharma, Aerospace, Meteo
... Scenario: multiple data and service providers for genetic research
... 2nd scenario: distribute data...
... 3rd scenario: collaboration
... Slide 6: SRS (sequence retrieval system)
... Slide 7: SRS federation
... semantic broker for discovery, composition

<mnot> wonders what "WS-I+ compliant" means...

Changtao describes the architecture design

Changtao Qu: Slide 9 Bioinformatics ontology
... Slide 10: Annotation/Publication
... based on OWL-S and UDDI
... Slide 11: Matchmaking & Composition
... Slide 12: Workflow
... uses OWL-S + XScufl
... Slide 13: Conclusions

Amit Sheth: question about workflow languages

Changtao Qu: in the grid community the focus is different
... we may extend some workflow languages

Amit Sheth: there are a few other ontologies that have been developed. ... semantics of composition could be limited ?

Changtao Qu: semantic composition should depend on user requirements
... I don't think we can do automatic composition
... it is too complicated

Amit Sheth: I agree it could be a dream
... if a lot of semantics was captured, it might help in composition

Duncan Hull: in a research environment, ontologies should be public

Changtao Qu: the project is commercial (SRS). It cannot be shared

Amit Sheth: I myself have been in the past designing open source ontologies (SWETO...)
... there is also public source tool to create ontologies (GLYCO)
... developped explicitly for performance experimentations

???: ??? syntactically interoperable? [was a question about logic DL/F-logic]

Describing Web Services for user-oriented retrieval - Duncan Hull

Duncan Hull: Slide 2 (outline)
... Slide 3: graph showing growth of gen data
... Slide 4: graph showing growth of the number of data bases
... Slide 5: we need to integrate a lot of sources to understand
... Slide 6: solutions for "integration"
... from cut-and-paste to Web services
... easier to exploit all data
... Slide 7: problems
... Slide 8: very crude description of the data (string)
... Slide 9: need for annotation not satisfied, so built an ontology
... description by annotation
... Slide 10: example
... (accessor-mediator)
... Slide 11: myGrid annotation tool
... does not use pre/post conditions or goals
... "semantic type" has to be detailed
... Slide 12...
... Slide 13: "wishlist"
... describe the minimum
... evaluate benefits (otherwise people won't annotate)
... scalability
... Slide 14 (wishlist cont'd)
... automation in longer term
... the workflow represent an experiment in bio
... adding mediators way break the workflow
... discovery (last)
... Slide 15: conclusions
... WSMO and OWL-S don't describe the highest priority stuff to be used in biotech
... RDFS/RDQL is enough for now
... need ontologies for _large_ numbers of services

David Martin: Can you say more about your last statement about RDFS/RDQL

Duncan Hull: we can have the minimum information useful to the users for those particular tasks

John Domingue: [....]

Duncan Hull: stability: a lot of web services will not be there in 10 yrs
... size matters in a sense that you need a good way to describe the WS
... not practical to go into a high level of details

John Domingue: if you need to change just one attribute, it scales up

Duncan Hull: you have a lot of services but it's not very clear about what they do (i.e. about what you are describing!)
... in a lot of services we don't have a lot of metadata in the wsdl


Semantic web and pan-European eGovernment services - Peter Brown

Peter Brown: I am not going to talk in the SW language
... will present 3 use cases
... presents background: growth of online services provided by public administrations
... I'm agnostic in terms of what direction we should follow
... it seems very premature to select a technology
... EU specificities (> 20 languages, 25 countries)
... result in differences in terminologies (wrong assumptions on equivalences)
... 1st Use Case: Administrations
... find any service (location/technology infrastructure/... )
... how to build a service description: data model, language, not important?
... service registry ?
... Components? service quality? limitations?
... 2nd Use Case: Businesses
... lots of paper-based procedures
... for volume of services of millions Euros
... requirements = location and language independant
... authentication, confidentiality
... 3rd Use Case: citizens
... eGovernance
... User moving inside EC
...requirements: high level of authentication/identification
... implicit semantics in what is required to buy a house or pay taxes

Rick Hull: could you explain more about the 2nd case?
... by nature in the area services have side effects
... it is not just a notion of workflow
... (much longer duration)

Peter Brown: since last year Austrian gov recognized electronic versions of admin papers as 'having the same value as an original'
... if you go to Italy with your birth certificate on a USB key, you won't go very far
... different policy rules across the states
... second point in transactions is persistence
... in terms of personal data I'll be in favor of personal data submitted in agency A of a gov would be asked to me by agency B if needed

John Domingue: [???]

Peter Brown: there is a proxy for the user
= an infrastructure fulfilling the auth technology requirements

Peter Brown: The bank is a proxy for your money - with a full audit trail

John Domingue: do you have a time scale (when countries will all have this)

Peter Brown: deadline in 2010 (?), but for individuals acting as e.g. CIO of a company, it will come later


David Martin: This session (UCs) was aiming at gathering requirements
... Need/Benefit and Technical requirements

Amith Sheth: Which UC is money-making?
... do you think that complexity is something that would be addressed by a standardization process?

Fabien Gandon: there is a difference between stdardization and research. There are simple scenarios that can be done with simple technologies out there (ready to be used, real demand). That's what we should get out of the lab.

Peter Brown: if you imagine a scenario for which the service providers can resolve and the SW can model the needs, there is a lot of money involved

Peter Brown: Anyone who ignores eProcurement needs to wake up!

<AmitSheth> Uses cases need to reflect current industry challenges demonstrating what would semantics buy-- such cases would probably find more excitement among industry partners/members

Todd Pehle: Relevant GIS info can save time in times of crisis

<AmitSheth> presentations observed complexity of current approaches-- presumably because they try to address the problem more comprehensively-- all aspects of web service/process lifecyclel lower hanging fruit and more targeted aspect could be initial items of investigations and considerations

Todd Pehle: Our Use Case is mostly about saving time (you would find the data anyway, but there is potential to save time).

Rick Hull: It is important to see the benefits beyond the use cases we have seen
... search and discovery, that you can't do with just a WSDL description
... link from WSDL, MEPs, etc. to the effects on the world

Fabien Gandon (INRIA): I want to reiterate that we need typed I/O and typed services

Harry Halpin: my question to the UC is "are you happy with the WS as they are?"
... (we see a lot of critics against the SW).
... What is you can't do without the SW? how does SW simplify

<AmitSheth> whatever we do cannot be revolutionary, it needs to be evolutionary, or there will be no takers

Axel Polleres: UDDI is not usable, give us anything usable, we'll be more than happy

Paul GianGarra: wrt security, lots of things in common with the military

Harry Halpin: We need web-scale identity, composition (through reasoning), and data integration.

<AmitSheth> public registries - UBRs --are not usable, but private ones have been used extensively; a question is whether with extensibility elements we can make it more useful-- the who industry classification idea went the wrong way-- brittle

Paul Giangarra: one of the things done today is the difference between extra-services and intra-services
... you keep data on intra-web
... how security plays on SWS?
... how information flows is critical to the success/failure

???: we see always very simple services in examples. I propose "managing complexity" as a requirement rather than saving time or compositionality
... Hiding complexity is an end-user benefit

Peter Brown: like in Google the ranking may not satisfy me, the complexity is related to the user's needs
... it is a benefit if you can expose the complexity and negotiate the solution

Ken Laskey: somebody at a WS conference would have the same list (of needs).

Sheila: certainly managing complexity needs semantics
... most of us make our travel plans and the web and we spend a lot of time gathering info and then matching to our needs/prefs

Ken Laskey: yes, so what?

Sheila: understanding what you need to use services, compose them... It's not just about ontology integration (what's the service angle).

Amit Sheth: the hardest thing to develop real world applications is to map the workflow to the data bits
... being able to publish a service, better find a service are the low hanging fruits (to me)

Peter Brown: isn't this semantic interop?

Amit Sheth: we have to be more precise

Rama Akkiraju: it would help developers build solutions

Benjamin Grosof: I agree that security is an important area

<AmitSheth> mapping (needed to achieve composition) and publications seem to be among the near term challenges/requirements

... for access, confidentiality...
... there is a need to a kind of access to the sw info [out of topic discussion about RULEML...]

???, Cefriel: [ontologies debate]

Katia Sycara: I don't see what is wrong with saying this is my ontology
... if you agree with my ontology, you can use it, if no, put your own ontology at a URI

Katia: Remember - "a little semantics goes a long way"

Bijan Parsia: is there anyone from a SW startup?
... is there anyone from a company developping a million dollar SW product
... is there anyone in the company that could spend.... is planning to spend a million bucks on a semweb product?

Julie DeMeester (MITRE): we have a number of project driven by customer projects
... but we can't disclose our plans because they're not public

Bijan: I'll be watching you all next year :)
... it seems to me ... why are we discussing about standardization?

Dominik Kuropka: we should make a 3rd slide to discuss risks and challenges

[David is starting a 3rd slide]

Dominik: can we get high investment pay back?

<bijan> To address sheila's point, do any of us *know* of any such companies, etc. etc.

<sbattle> Bijan's comments were made in "the spirit of hope" :)

<bijan> Alas, my hope was not realized :)

Alois Reitbauer: automation of tasks is important since scenarios require more and more complexity
... applications are evolving, need reconfiguration, adaptation
... don't require too specialized developpers

<TerryP> KISS - Keep it semantically simple :)

... mapping from workflow to data: is it really a benefit???

Paul Giangarra: many of us work for companies who spend lots of money in WS...
... there is a bullet about discovery, it says "in crisis situations"
... it is probably a different UC
... what kind of discovery? location? needs?
... make "discovery" more precise
... if SWS are going to be successful, will WS-I put them in a profile?

Hugo Haas: Bijan, W3C have vendor org. members, but not only (you are not a vendor)
... you are so pessimistic about the C&C workshop because 2 solutions were presented
... Paul, WS-I was created because SOAP1.1 and WSDL1.1 were out and proprietary and people did not know how to fix the problems.
... SOAP 1.2 had XML Protocol WG to fix it

Julie: I would every company here has prospects, may not have a product to deliver in 2006 but is working on something
... we are manded for profiles when developing WS specs
... in the DoD we have a big investment in XML, XML tech, and now in WS techs
... what do I do with my WSDL that I have today?
... question is to get to the next steps


Minutes formatted by David Booth's scribe.perl version 1.126 (CVS log)
$Date: 2005/06/14 16:21:02 $