9:15 - WSMF - Dieter Fensel

machine processable semantics:
assume some pieces will break
  a decomposable system allows the machine to try other options
  no human requirements

ontology tech to manage didefferent docs

:

strong decoupling via interface description
strong mediation services

WSMF:
ontologies
goal repositories
description
mediators - should be web services again but is very specialized

Ibrow - intelligent brokers that are able to distributively configure reusable components into knowledge systems through the World-Wide Web
SWWS - Semantic Web enabled Web Services

invisible web
  best tools are the one you never see

Q: similarity betweens Semweb services and what we used to call agents
   lots of talk about frameworks and archs but few examples -- funders should ask for examples
A: yes there are strong links between agents and Wemweb services, (i usually hide from industry), but it doens't make sense to think that every web service is an agent.
   yet-another-framework: all players agree that they suffer from a lack of standardization above SOAP. additional layers needed before it is really workable. most of the money is in enterprise and large organizations. for academia i address these enterprise issues.

Q: how to establish trust relationships between services on the web
A: we view this as a non-functional properties -- it goes into QoS. tricky when following the chain of components

Q: component based software development was a great legend of the 90s. why do you push it further and what will make it a success.
A: we are allowing more parameterisations. we have had success in automatic programming when the components used dervitaves of the same ontology.

[show dieter CC/PP re mediators - done]

----
11:15 - DAML-S - David Martin

WSDL provides most of the answers we need for grounding.

Using extensibility to use typing systems other than XSD (which is the default).

[Q: Is it possible to compose a process by examining the inpu
If I publish a service with an input that I can't imagine being the product of some service, 
Is there a convenient route from a precondition or input the the advertiser did not view as ]

Q: is there a semantic difference between two web services, how do i integrate them
A: the semantics would be provided by a human developer and the diff between the services would be made clear in the preconditions or inputs.

----
11:35 - storage and querying middlewear for the semantic web
*** presentation is different than paper in the proceedings

Q: Is it [Sesame] a complete RQL impl?
A: 99%

Q: Jeremy: what are the different tensions between storage and inference.
A: schema-specific methods in the API.

----
11:55 - Mini-experiment in Semantic Annotation

visual annotations system

[talk to Lilo about ICONCLASS]

Q: was RDF schema sufficient?
A: extremely flexible but needed an equiv

Q: comment on this vs. already deployed and used systems [in industry, not like annotea]
A: they liked it. it was easier and led them toward what they were needed.
 need to combine with some atomic processing [?]

Q: where is Wordnet used?
A: the ontology selection you saw integrated wordnet.

will now do more ontolyg management on relatedness predicates

----
12:10 SymOntoX

fields:
  spec: isa
  decomp: part
  predication: concept or instance to value
  similarity
  relatedness: tourist --staysIn-> hotel where "staysIn" is the relatedness

[systems enables classification of art pieces by keying into and presenting onotology terms.
 use enters "bed" and selects meaning of a bedroom kind of bed.]

[storage: RDB with a few dublic core fields and an relatedness for the rest of it.
goofy as it was not connected with RDF at all.]

Q: Harold: You only use binary relations. Did you consider RDF?
A: [hem haw]

Q: Will you work more on interfaces?

[Chutiporn's P&P -- discuss CC/PP - done]

----
14:30 - Problems With the Semantic Web - Peter Patel-Schneider

re meaning from HTML:
  they are meant to be little more than pleasing to the eye  - or -
  content providers change their data format to prevent screen scraping



[Q: XML has no presumption that you understand the meaning of an element without understanding the ALL of its parents, meaning: understanding gets traced from the root node.
]

Q: (looks like nested relational model.) what about RDFS?
A: let's get rid of it. i don't think it hits the right point between XML and an ont lang.

Q: Ora: (i like this a lot.) What if you took something else that's built on top of XML, like XHTML?
A: maybe if people mark up HTML to have semantic import. Also maybe topic maps.

Q: Jeremy: XML meaning, XML community have PSVI, canon, ... need a schema-specific transformation. need the transformation lang.
A: agreed, very viable competitor
Q:         closed/open model assumption. RDF not entitled to make closed assumptions but purchase order example may have extra properties as it doesn't say that it doesn't.

Q: HP: do you preserve the graph merge opperation?
A: yes.


----
15:00 - Taking the RDF Model theory out for a spin - Ora

----
15:30 - formal model for topic maps

Q: Harold: where are the roles?
A: just need to defined the relation between the topics.

[pester to look at mapping]

----
15:50 - participating in the Sem Web too difficult?

Q: Harold: would ti be possible to generate different meta-data for different audiences?
A: Yes, but also true of XSLT.

Q: You usd UML, was it sufficient?
A: OCL seemed to be effective.

----
17:15 - Graph Matching - Jeremy

Q: Jim: have you considered this applied to query?
A: graphs with variables are essentially bnodes. but subgraph isomorphism, known to be NP, must harder

Q: Harold: something about undirected graphs
A: since undirected graphs can be rendered in directed graphs, these methods apply and we should be using them.

----
Trust

military intelligence trust project

[they may use annotea, though maybe just an example of annotation]

Q: you view of trust seems absolute. what about contextual trust?
A: i expect capturing the opinions of diverse users is key. at least capture the data for now.

Q: Harold: did you consider inheritence of trust, eg bank, credit card, ...
A: no. not yet.
Q: propagation of trust. truth maintainance technology.
A: what will we do when some piece of our conclusion is based on flakey data, do we trust it?

----
Querying the Semantic Web - a Formal Approach
Ian Horroks

Q: Jeremy: 
A: yes, you can do 
Q: Jeremy: most query systems are assume open world.
A: if you ask for everyone whe is not a member of some group, it won't give you everyone whom you haven't bothered to say isn't a member of the group.
   in that respect, it may give you different results than you expect.

----
RDF Query API
Andy

Q: What does Joseki
A: Go term covering operations where the consequences are pretty well worked out

Q: you want add and delete, would the ... people's quer-based operations be sufficient?
A: there are consistency problems. if you pass a query to select the deletes, it may select rather more than you expected without telling you. May need locking.
   the locking problems creep in as the network gets larger.