See also: IRC log
<scribe> Scribe: Jeff Waters
<scribe> ScribeNick: jeffw
jeffw: Hi, Piotr and Eva!
<eblomqvi> hi
jeffw: We have Marion on the telephone as well.
<eblomqvi> I am calling in now
<piotr_nowara> hi
jeffw: This is a fun time for our
incubator, since we have a nice position paper which Eva has
wonderfully edited and prepared for the Workshop on Ontology
Patterns for ISWC 2010.
... Piotr, will you be able to join us on the phone as
well?
<piotr_nowara> i'm trying to make a call
jeffw: Thank you Eva for you edits and fixes for the work.
<piotr_nowara> it seems that the french line doesn't work
eblomqvi: Jeff, will you be able to provide some input?
jeffw: Yes, I'll be able to provide the inputs you requested.
<piotr_nowara> i've joined the teleconference
jeffw: I thought the reviewers liked the topic, and appreciated the readability, but in some ways wanted more.
eblomqvi: I had a related topic,
regarding the agenda, and there was another opportunity for a
practical exercise in the workshop trying to produce some
patterns.
... We have a call for proposals for things to work at the
workshop and I suggested we had some modeling problem to submit
to the workshop
... I found an old email where Jeff listed some ideas that
we've encountered that might be worth submitting, like the
metrics issues and a few others.
... The deadline is also tomorrow, if you would like to proceed
it's not a huge effort to submit to the patterns portal as a
suggestion, if it happens to get selected, it could be helpful
and nice opportunity to see what would come out of it.
... In case you are interested, I could try to describe one or
two of these problems and we could see if we could submit by
tomorrow.
marion: That sounds like a great idea. I have a question about the page limit, is it 4 or 5?
eblomqvi: it is 4, so I will still need to cut it down tomorrow afternoon, so the deadline I gave you was a little early.
marion: I could look through and make suggestions for cutting.
jeffw: Yes, and I agree that we should submit some problem, there is
jeffw: so we've been discussing that, did anyone else have any comments?
marion: I have some suggestions, but it would help if I could work from a version I could edit
eblomqvi: You can edit in a plain
text editor, the latex file, you'll see some strange sequences
but you can at least edit the plain text
... You should be able to edit with notepad fine.
... We can use Word next time, I just started writing while
people were on vacation, but if we can survive for now, we'll
figure it out better next time.
jeffw: ok, one pattern that
occurs in the options is an ordered list, so one issue is how
best to represent an ordered list
... Is there a problem with using rdf:Sequence for an ordered
list? I know there are other solutions for modeling ordered
lists.
eblomqvi: It's similar to problem
of named graphs, you want to talk about it and be on the
modeling level rather than the data level, I think I sent you
this link once before
... but I will post again, so it's one of the solutions for how
to model this in OWL. So I will post this link for you.
<eblomqvi> http://swan.mindinformatics.org/spec/1.2/collections.html
jeffw: Just the existence of other solutions suggests there may be problems with the rdf solutions.
eblomqvi: rdf solutions are just talking about the data, but if you want to say things about the constructs of the list that's the main difference
jeffw: ok, I see that, that insight will help me think about it.
<eblomqvi> the link I posted above is one suggested solution for modelling sets, bags and lists
jeffw: Piotr, do you have a preferred solution for ordering?
piotr_nowara: I found this solution that Eva found, but I don't have a preferred solution.
jeffw: named graph issue, I wanted to use the concept of question as a component of a decision, so I found the quads and named graphs as the better solution, any thoughts?
eblomqvi: re provenance, it is the same sort of problem, it seems like people are heading in this direction as opposed to reification, we are going in the direction others are
piotr_nowara: modeling questions, questions can be modeled as a statement, so you're statement with variable idea,
jeffw: For how to model this, I was thinking about a variable class, then subclasses like who, what, where, when and then perhaps subclasses like whatComputer, whatCity
marion: consider upper ontologies, that would be a good idea
eblomqvi: There are different types of question, if you ask who, it should be of type Person, if where, then of type Location, so we could look at question/answering literature, maybe I could have a look and pull some links
jeffw: Yes, I'd very much anything you can find and it could be posted to the wiki or whatever. That would help me.
marion: we need to understand what and how and why we are answering the question
eblomqvi: if you really want to do question/answering, we won't necessarily get the correct answer, like weighing options, we are not necessarily finding THE answer as much as allowing for uncertainty
marion: You have to allow counter-factuals or else you can't consider the options appropriately.
jeffw: Can you give me an example?
marion: it's a what-if question, it poses as a proposition, it suggests a situation that isn't true and then reasoning on that to see what would be if it were true
jeffw: yes, I like your idea, and it goes beyond what I was describing, i wanted a way to point to a machine understandable version of a question and you are suggesting going beyond so that it points to a machine understandable version of a question that can help you actually come up with a solution
marion: you could consider the
outcome, have the computer rank order based on pre-set,
etc.
... What could exist now v. what could exist if other things
occurred and there is secondary issues like cognition under
stress and it seems to me this is doable
... if you could indicate this is a question, then if no known
domains refer to it, then you could consider how to
incorporate
piotr_nowara: I'd like to discuss
the issue about the question, I'd like to suggest introducing a
boolean variable that some questions like Should I buy this
kind of computer where the answer is yes/no
... so we should consider this type of variable.
eblomqvi: for the other core
decisions model, I promised to answer the e-mail and I haven't,
but I got some of the discussion with some interesting points,
the notion of a situation could be used to model criteria etc.
and I did'nt have a chance to think hard on this
... but I should have more time in the coming week, piotr would
you like to recapture your thoughts on criteria?
piotr_nowara: I need a couple
days to explore my criteria pattern and how they are related,
to summarize, I think criterion is a point of reference by
which we can evaluate, and it is determined by something, like
a regulation or constraint that sets some criterion, a
criterion center
... but determining the domain .... (I think we lost the
connection)
eblomqvi: I am very interested to see your progress on this criterion pattern, I'd be happy to see it even if not the final version and we could learn and utilize
piotr_nowara: the requirements
determine which domain for which the criteria is valid, there
is a yes/no but there is the quality of SQL knowledge, so the
situation pattern could be used to express that domain
connection, should we introduce the broader concept of the
criterion
... or should we just do a simple pattern
eblomqvi: Piotr, I would suggest you do both, cause we could take a modular approach, you could do the basic pattern and in a more detailed version, you expand, would that work?
piotr_nowara: I like that idea and I'll let you know when I have some results
jeffw: The concept of the tool is
that you might begin with a question, like "What city should I
vacation in?" and you could have some keywords, like "vacation"
and "city", and these keywords could drive a search for
relevant linked open data sets.
... Once those datasets are listed, you could browse them to
see which one or more might be relevant and you pick one (or
more)
... Then the items in that dataset become your options, for
example "cities", and the properties of that dataset, e.g.
population, number of museums, etc., become the potential
criteria.
... you then pick the properties you want for the criteria, how
you want to weight them and any min/max threshold, such as
population < 300,000 with a weight of 1.5
... then the criteria are applied to the options both to filter
out options which don't fit the boundary conditions, but also
to order the options based on the criteria and its weight,
assuming the criteria can all be quantified and normalized for
comparative purposes.
... Then you make your decision based on the reordered options.
All of this is captured in our decision format including the
amount of time spent in which states of the decision
process.
... That's the concept and I'll ask our tool developers to see
if they can join us next time to discuss this in more
detail.
... Thanks everyone for calling in. Goodbye!
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Found Scribe: Jeff Waters Found ScribeNick: jeffw WARNING: No "Present: ... " found! Possibly Present: ScribeNick eblomqvi jeffw marion piotr_nowara You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy Agenda: http://www.w3.org/2005/Incubator/decision/wiki/Decision_Mtg_14_Agenda Got date from IRC log name: 30 Sep 2010 Guessing minutes URL: http://www.w3.org/2010/09/30-decision-xg-minutes.html People with action items:[End of scribe.perl diagnostic output]