See also: IRC log
-> http://www.w3.org/2008/Talks/1209-W3C-EnergyOandG-Workshop/Intro-W3C-OandG-bratt.pdf Steve's slides
Onno Paap, fluor corp: is there an exponential grow with the number of companies in the hcls group?
steve: no, it increases but not exponentially. They said there is no one place where people can really get together, we know that w3c does not have any particular agenda..
onno: our goal is to go for a special group for process plan, what is the best way to start?
steve: they started with very broad goals in
hcls, but after their workshop they set up task forces
... they need the right people to set that up
... this is something that has to be done
roger: chevron has been a w3c member since 2000,
I was at the start at the hcls
... one of the advantages is also that one can get to know some of the most
extraordinary people in the industry
... it is a huge advantage
-> http://www.w3.org/2008/12/ogws-slides/Crompton.pdf Jim Crompton's slides
<steve> did Jim say 6500 feet of water?
<steve> impressive
<kendall> yes
<kendall> with some linked component at 7k
<LeeF> jim: challenge is to bring data we can produce to the analytic people/software in a timely fashion
Richard Sears, Shell/MIT: this is not a chevron problem alone, i discussed at shell we discussed that the knowledge is jsut huge, we know that the data is there but we cannot find it.. shell people said they are interviewing retireees to know what they know
jeremy: you speak about increase data quallity,
what I think is more important is the quality of the metadata
... then you can integrate it in proper fashion
bertrand: in the 70's the industry began to build
up their own database
... we then went tot he 80's we removed the databases to commercial systems
... we saw the first companies selling products
... but at the end of the 80's all looked different and that created
problems
... we are in a risk sharing industry, when we build platforms we want to
share the risks, but we then have to share data
... this is not out of love but out of necessity
... there were two ways to happen
... some companies tried to foster their view, some others wanted to let it do
together
... in the early 90's a set of organizations were created to build common data
models
... that did not work.
... we tried to build an esperanto for the oil field and and the fate was one
of esperanto. it did not work
... what we did was to look at the problem at a different manner.
... we tried to to find ways to speak together. focusing not on common model
but on common model of exchanging
... we created something that was a success. the idea of exchanging through
xml was successful
... what we have a good conceptualization for what exploration is today
<Robin> Joined the POSC organization that had a budget of 8 million annually. However with this annual budget conclusion was that there was no ROI
bertrand: we have a good understanding what we
are talking about
... we are capable of going to the next level, to express the value of that
understanding
... that is what we are tallking about now
... maybe we are the 80's to make a huge mistake or the 90's to really do what
we need
-> http://www.w3.org/2008/12/ogws-slides/Design_and_Implementation_of_a_Semantic_Web_Solution.pdf slides
<Robin> IAM: Integrated Asset Management
<Robin> IAM contains a "Metacatalog" which is an OWL triple store
<Robin> IAM development started in late 2004 and target to complete late 2008
<Robin> Use of upper ontologies have proved to be hard
<Robin> During the project it was necessary to change/extend the ontology
<Robin> These changes resulted in change management problems
<Robin> The expantion of the ontology resulted in "dirty" queries which required the maintenance of SPARQL
Fred van Home: as the ontology evolved did it break the queries a lot, and you have added concept that break the queries
amol when you have to break the ontology because that is what the evolution requires, than this happens, then the semantics may become different
scribe: it is hard not to break queries when you evolve an ontology
fred van home: so you did change the meaning of previous terms. you use a term 'well' to keep it as a superclass, but you have to introduce new subclasses
Ram: it does not necessarily happen a lot, but the point is to know when and how it happens
amol: it is also a question of style to define
ontologies
... we approached to follow as requirements evolve
robin benjamins: that is the way i see it on th ebusinees side, one goes bottom up
ram: that means we have to go a pragmatic way in ontology design
-> http://www.w3.org/2008/12/ogws-slides/cambridge-semantics.pdf slides
<Robin> Emphasizes the modeling of a domain – information modeling free of technology stores like relational database, XML, etc
<Robin> a. The area where we can extract more automation is through the integration of silo data. This takes a considerable expertise and resources. Solution is to use “people power”
<Robin> Everyone loves Excel. It is our most dominant application/data store
<Robin> Cambridge Semantics vision is to tie the model with Excel, and machine processing
<Robin> Approach: target PRODML and WITSML as a useful machine representation
<Robin> Goal is to transform content in Excel to PRODML
<Robin> Approach is to use a program/macro called Anzo in Excel to perform transformation
<Robin> Anzo allows for mapping of Excel cells to an ontology
<Robin> Anzo uses templates to apply against similarly structured Excel to reuse ontology mappings
<Robin> Anso Expose is another application that allows you to work with RDF data. Loads in data from a Anzo mapped Excel
<Robin> Anzo Expose provides visualization to support analysis
<Robin> Anzo Expose an convert content to PRODML
<Robin> Anzo Expose supports mashups between content from Excel sources with Google maps, etc
bertrand: nobody want to use these kind of tools... you ask the human to create computer perspective...
lee: we tried to automate what we could and these
things can be done autoamtically
... but people are already doing this but doing it manually
... thousands of hours are already used to do this, this is an improvement
over there
roger: when we saw that is that if somebody
changes a spreadsheet, all changes are propagated all around
... the spreadsheet becomes part of the business information
neil McNaughton: the problem is that people do not really use spreadsheet well, can you go back to produce better spreadsheets?
lee: we plan to do that, create ranges, better formulae, etc, but it is also difficult to do it
-> http://www.w3.org/2008/12/ogws-slides/Chevron_Position_Paper.pdf slides
<Robin> Chevron is on path to optimizing the entire information chain
<Robin> The number one role of Semantic Web is data exchange of information between applications
<Robin> Pilot: ITC/ETC – originally an internal collaborative effort. To develop a way for Chevron to search its documents
<Robin> Focus was on a subset of technical data. Began be identify the enties in the data. Then developed ontology. Created an RDF store. Discovered that RDF as a standard was solid. Learned during the project how to develop ontologies. They determined that developing ontologies in a modular method was best. To define the universal proved to be out of scope
<Robin> Used TopBraid application to develop ontologies
<Robin> The resulting solution/approach was discovered to work on other use cases
<steve> Is this a useful doc re: URI schemes?
<steve> http://www.w3.org/TR/2008/NOTE-cooluris-20081203/
<steve> ... or is there another one from HCLS IG?
<LeeF> It's ok, but I'm not sure it really addresses the types of challenges that we're talking about here
<Roger> I think that this document would have been very helpful when they had their biggest troubles.
<Roger> But I don't think it was published at the time.
<Roger> Well, at least somewhat helpful.
<Roger> They were really struggling at first.
<LeeF> :-) I just meant that it doesn't deal so much with how to get stable URIs, I don't think
<Robin> Key Challenges: ontology development is complex, needs to be hand built, expertise is hard to acquire
<kendall> so what are the challenges? it wasn't clear from what he said (well, clear to me)
<LeeF> Not clear to me either - my guess was that it was about how to go from file (data) to stable identifier in the face of changing contents - but that's largely me projecting my thoughts onto what he said :)
<Robin> SPE is setting up an open oil field ontology
<rdecarlo73> re:uris - I agree, stable URI is a big issue for environments where ownership changes over time
<Roger> They kept making URI's based on schemes that they regretted later.
<kendall> Sounds more like a modeling than URI issue per se, but the boundaries can be fuzzy.
<ChevronRoger> Well, it was the NAMES of the URI's, not the concepts being modeled our their relationships.
<ChevronRoger> I'm not quite sure exactly why it gave them such fits, I'm just reporting it.
<ChevronRoger> YOu might ask Mario.
<kendall> but that's a modeling issue in some sense, since URIs are supposed to be opaque (the URI string itself, I mean) and the data about the resources they (opaquely) identify should be in some representation (RDF, etc), which gets us to modeling per se
<ChevronRoger> I don't think they were opaque enough, particularly at the beginning.
<kendall> that can be a problem :)
<kendall> thanks for clarifying
<ChevronRoger> As I said, I think that doc would have helped them.
robin benjamins: we see this as a long term investment. At the
same time i have to deliver annual value
... the strategy is a long term vision and start to bring in your current
needs
-> http://www.w3.org/2008/12/ogws-slides/ENI_Dante.pdf slides
<Robin> Web 3.0 is putting Web 1.0 on its head
<Robin> Web 1.0 is single point of content with mass consumption
<Robin> Web 2.0 is a social web
<Robin> Web 3.0 mass production and pinpoint consumption
??: how do you handle modifiers
(scribe lost track, sorry...)
lee: do they use this with a human layer in that, or is the result good enough out of the box?
The answer is: there is always a human being (linguist) who takes account of the uniqueness of terms in the industry
<LeeF> ???: is there a way to use the information you get from this process to feed back to content authors so that they can help produce content that can be processed with greater precision?
<robert> Neil MacNaughton -> Neil McNaughton
<LeeF> Brooke: in our case, there is a feedback loop ... there is a web interface to require a concept rather than a keyword - think that is a reasonable sweet spot to guide people
<ivan> Kendall's slides
kendall: summarizing houston weather, making the business case for Semantic Web, and updating software on presentation laptop without missing a beat
<FrankC> Finally!
<FrankC> Integration and analysis feedback loop
<FrankC> more integration --> more analysis --> more integration...
<FrankC> Sounds familiar in O&G IT
Onno Paap: What do I get from semantic web technologies that's better than what we're doing today for things like skills databases with Excel?
Kendall: it was more than that - incorporated LDAP, skills definitions, technical publication repositories, ... Was between 8 and 10 different sources that needed to be integrated
May Hsu: we're encouraged to reuse ontologies - do you start from scratch or do you use existing ontologies in a new project (like NASA)?
kendall: we did a bit of both - where we could we, reused or extended existing ontologies (e.g. extended FOAF for people data). Used an XML schema (converted to OWL) for competency
-> http://www.w3.org/2008/11/ogws-papers/carroll.pdf Jeremy Carroll / TopQuadrant position paper
<Robin> Key Roles: Missed an important role on slide 6 - The Leader
following slides, demonstration of TopBraid Composer
D2RQ + similar for XML schema to map coordinate database & WITSML data to triples
<Robin> ontology source demonstrated in TopBraid Composer: http://www.epsg.org & http://www.witsml.org
bertrand: if you had taken the data and built the ontology from scratch, what would the difference be?
Chip: there would be a much richer class
structure, rather than the flat generated one - you might have started with an
upper ontology and built the model from
... also would have thought more about ways to force interoperability in the
model
Jeremy: another key difference is cost
-> http://www.w3.org/2008/12/ogws-slides/Bentley.pdf slides
<FrankC> We would like to collect discussion topics for tomorrow's panel discussion
-> http://www.w3.org/2008/12/ogws-slides/Fluor.pdf slides
<mario> ISO 15926 Interoperability standard
Neutral layer used for data integration -->open layer?
bubble chart for the Uso classes
and their relationships
<mario> FIATECH choose to go with ISO 15926 as a standard
Interoperability workshop held in Houston May 2008 resulting in a number of vendors (Intergraph, Oracle, Adobe, Siemens, etc.) wanting to use ISO 15926
<mario> Enable data exchange between vendors that leverage the standard
ISO 15926 provides the ability to build common data models
<mario> Facade is a triple store
Lots of spreadsheets!
<LeeF> +1 spreadsheets
<LeeF> and +1 SPARQL
Legacy system models map to Facade --> you can the facade and map back to the system
<kendall> -1 to the claim that you never federate into a central store... there are lots of use cases where that's perfectly sane
ISO/TC67 Oil industry standards
<LeeF> -1 to Excel is out :-)
DNV paper: Wanted a simple, compliant interface
RDL has 20 thousands property classes
Template --> predicate
Template can be implicit with rule
Publishing SWRL rule in repository as well.
Translation of propriatary databases through an ontology DB / templates
Standardize data exchange in template format
Knowledge representing data models with rules being described with templates
<kendall> I wonder if they've looked at OWL 2's new features related to n-ary datatype predicates and some concrete domain reasoning that might clean up a lot of this template and datatype stuff
Yet another modeling technique
Probably not
<kendall> yes, that's my guess; but it might really simplify some of the stuff they've had to add-on to handle their modeling requirements. Which would be a win, presumably.
No joke
<kendall> well, i'm being circumspect about that on purpose since I don't really understand this templates thing *at all*. But maybe I'm the only one. :>
Part 7 template using RDF syntax but with yet another axiomatic definition
me neither
Ivan: Security and provenance issue for SW still have more work to be done
<kendall> ah, XACML gets a mention, nice
<kendall> but that's a major faux pas to mention an OASIS standard at a W3C event
<kendall> :)_
You hit the nail on its head!
<kendall> (for the record, I was joking and don't really think this is an actual problem!)