W3C

- DRAFT -

Decisions & Decision-Making Incubator Meeting 4

13 May 2010

Agenda

See also: IRC log

Attendees

Present
Regrets
Chair
jeff
Scribe
jeff

Contents


<scribe> scribe: jeff

<scribe> ScribeNick: jeff

Welcome to the May 13th, 2010 Meeting of the Decisions and Decision-Making Incubator Activity The agenda can be found at http://www.w3.org/2005/Incubator/decision/wiki/Decision_Mtg_4_Agenda Let's begin with Topic: Status

Status

Hi, Eva

<eblomqvi> hi

Regarding status, we have outlined some initial use cases and we are beginning to apply our eXtreme Design (XD) approach to our decision components that we've initially identified. In order to do this, it is important to consider and become familiar with relevant ontology design patterns. As we proceed, we also recognize the need to educate ourselves on the appropriate use of the tools (e.g. Neon), the use of the semantic standards (e.g. RDF, OWL

led through those standards (e.g. Open Linked Data) as subject matter and context for our decisions. A fruitful agenda for any of our meetings would include a mixture of both specific application of XD to selected decision components, as well as more general but focused educational items on the tools/patterns/standards/linkedData efforts. The latter can be fostered through our own efforts as well as invited participants willing to share their expe

So in short we are pursuing perhaps 3 parallel paths simultaneously: (1) eXtreme design of a decision representation; (2) education of ourselves and others who are new to these efforts (ontologies, semantic standards,extreme design, open linked data) of the significance and what it means and how to do it; and

(3) Application of the decision representation utilizing open linked data (including sample tools, e.g. Ekkli, mobile app, etc.). One near-term goal: I would like to use a decision representation format by end of June (about 3 more meetings)

Marion: Ontology work ties in with 6.1 ONR project and have synergy

Progress on Application of Transition Design Pattern for Information Flow Use Case

The short answer is that I haven't had a chance to move this forward much yet, so I will summarize my thinking and my plan. Eva's colleague, Aldo, has prepared a nice ontology design pattern for Transitions, see http://ontologydesignpatterns.org/wiki/Submissions:Transition. This is one of many useful patterns which CNR has provided on their repository website. One goal of eXtreme Design is to reuse these patterns when the occur in specific ontology

flow use case, see http://www.w3.org/2005/Incubator/decision/wiki/Use_Case_Measuring_Info_Flow, which identifies a need for measuring the time spent in different decision states related to information flow, states such as "Gathering Information", "Analyzing Information", "Making Decision", "Preparing Decision Product", "Communicating Decision", etc. The goal is to have a mechanism to track how much time is spent in what phases of gathering and us

tansferring information in the decision process. It seemed like this might be an opportunity to reuse the Transition pattern, since both the pattern and our use case involve "states", "transitions", "time intervals", etc.

The next steps are to load the ontology into Neon, specialize (for example subclass certain components and add any special items needed to apply this pattern for decisions), and then unit test via SPARQL queries. I want to make some significant progress on this as well as a "core" decision component use case for our next meeting.

Overview of Selected Design Patterns Relevant for Decision-Making

Eva: A couple commnents. Last time we talked, you had discovered a bug in the XD plug-in, and now it is fixed, so please update to the latest version of plug-in
... Actually, we found some small errors in the transition pattern which may be why it wasn't loading in Protege
... I was fixing those in the Pattern, so I need to switch the file, so by tomorrow, the Transition pattern will be fixed

Thanks for doing all that.

Eva: This is the good part, when someone starts to use it, you see some problems.
... Then we wanted to look at a few of the patterns in the portal, maybe 70 content patterns, I didn't put the link to the page but I will put in the chat:

<eblomqvi> http://ontologydesignpatterns.org/wiki/Submissions:ContentOPs

Eva: There is a long list of patterns and the first thing you can look at is the domain and the intent, the latter is what is the purpose of the pattern

And it describes requirements and the domain gives you some hints for what the patterns were designed for, such as fisheries or biology, for decisions we may want something more general

Eva: So basically I wanted to mention that patterns are in a way ordered, there is a small set of general and basic patterns that are reused by more complex ones, like Transition
... So this is also one thing to consider. Should I use the general small components, such as TimeInterval, or take the bigger patterns, if the bigger fits well then use it, but if not use the smaller
... In the agenda, I listed some of the basic patterns that are the most common ones you need when you model.
... First is situation, you can find it at the link above
... Graphical view, competency questions, elements of the pattern, scenario and so on
... Situation contains the one concept with two properties, hasSetting and settingFor. What this pattern does is n-ary relation.
... If you have several things that are involved that you need to connect together. So tutorial example is theatre with Shakespeare plays and we had theatre festival that connected plays to times to places
... The situation is a quite general n-ary relation. You can reuse it and specialize it by making subclasses, such as TheatreFestival and have subproperties such as festivalTime, festivalPlace and so on
... so you create your own version. You can think do I need to import and reuse or do you just use the solution the pattern suggests?
... An advantage of import and reuse is that others can see the connection and can use it
... TimeInterval is another common pattern. We can look at the time interval page. You can see that it specifies one concept, the time interval, with three properties
... IntervalStartDate and IntervalStopDate
... What you don't see is that the two intervals are under one intervalDate, so you use the specializations,

I was confused about the cardinality about some of the Transition components

Eva: Let's jump to transition pattern to address your question. Overall idea is to represent transitions between states. We have a process, the overall decision in our case
... The process contains a set of transitions, which take you from one state to another state, and each are triggered from one state to another state, and there is an object involved
... Such as preparing a decision and having the decision ready, then the object is the product generated in the transition
... So you mention the times, you can see that you have the atTime property which is a minimum of 3 intervals, and it is defined more generally
... It has 3 subproperties, here is the mistake in the current version of the Transition pattern, the subproperties haveEventAtTime, haveFinalStateAtTime, and haveInitialStateAtTime, have a range of TimingSituation but should be TimeInterval
... AtTime has range TimeInterval, so subproperties should have the same range or more specific, so we fixed that
... Why 3 times? It is because we want one time interval for the event for the event, the beginning state and the final state.
... There is a complex component in the pattern, these times are now connected to the transition, but you also have the timing participation, the states can be connected directly to a time interval
... You don't have property chains in OWL 1, so you had no way to automatically extract haveInitialStateAtTime from properties from state
... You have both versions, cardinalities in OWL are not constraints, they are used for reasoning, so you can leave them unfilled

Jeff: hasBrother example, if you use domains and ranges, they are not constraints, they are used for inferencing

Eva: One way you can do this is for restrictions.

Marion: A more general way is to promote concept of sibling into the upper ontology so it can be used by both people and other animals, if desired, in which case you wouldn't need restrictions

Eva: You can define the domain as a more general class, but ...

Marion: Important to have a complete upper ontology so you can inherit without having to put alot of restrictions on

Eva: Good, yes, you can put general domain and range and it holds for anything, but the point is that you want to do some reasoning

Jeff: Another interesting cardinality for isSettingFor in Transition is minimum 8 items

Eva: The object involved, the process, the states, the time intervals
... I think we covered general idea of transition pattern, so we can try to apply it, I'll briefly mention a couple other patterns
... There is another pattern that may or may not be interesting, it is the control flow pattern, it is more about describing the plan of a process, whereas Transition describes what is going on
... ControlFlow is more about work flow, branches or paths one can take, so keep that in mind in case its useful
... Then there are also patterns called TimingParticipation, this is a useful pattern representing events at different times, we could model who participates in meetings, etc.
... A few small general patterns, there is the collection pattern which models collections of things and membership in collections
... You have also the sequence pattern which is used in Transition pattern, to say one state comes before another and you can use these patterns independently if you need to talk about sequence of things
... partOf is a transitive relation, so you can say the engine is partOf this car and so if you say something ispart of the engine, then it will be inferred to be part of the car
... Componency does something similar but allows for non-transitive relation. For example, you may not want to infer screw is a component of the car, so it keeps

Marion: It's called ...., to represent hierarchical substructures in ontology, x has y as opposed to x is y,
... like two ontologies side-by-side, one is for partOf and one for relationships of states of being, two different hierarchies in ontology

Overview of EDXL Emergency Management Standards and Impact on Decision-Making

Don and I both are familiar with projects related to information sharing in the Emergency Management. These efforts include alerting, hospital availability, resource allocation, and routing of emergency information. The efforts include the most immediate level of response,

information to and from first responders in the field and the chemical or other sensors they may carry, as well as the information to and from Emergency Operations Centers where higher level decisions are being made about evacuation, shelter, public information, etc. This is

only one of many potential application domains for our decision representation efforts, but it is a useful example application domain. We thought it might be useful to summarize some of the standards that have been developed or are being developed to support information

sharing in this domain. The eventual goal then is to become more familiar and address how this information might support or utilize our decision format. We'll do this quickly today, just covering the name of the standard, its purpose, the issues it's trying to address, and

an idea of how it might apply to our decisions work. I've provided some links which you can find from the agenda and the EDXL wiki page to some background information on the standards I'll mention, see

http://www.w3.org/2005/Incubator/decision/wiki/EDXL_Overview. Note that these are XML standards, and the use of semantic standards alone is an important consideration for us and part of our education efforts for those working on the current standards.

For example, how best can semantic standards add value to the current work.

Let's consider first the Common Alert Protocol (CAP). It's a simple (elegant) standard, the schema is only 2.5 pages with alot of white space. It includes components, like type of alert, confidence, priority, as well as title and description, and one of its primary

intended uses is for alerts to the public. The alerts could be for anything, hurricane, flood, evacuation, ... These alerts, for our purposes, could be considered the output from decision-making. They can also be inputs for further decision-making.

Next is the Distribution Element.. This standard is a wrapper to assist with the routing of message payloads and it contains components like senderRole, recipientRole, Keywords.

These components utilize a structure referred to as a ValueListURN, where the URN points to an externally managed list or set and the specified values come from that list.

This structure enables communities to use their own sets of terms and evolve more general, standardized sets. The architects of this work envisioned that the

external "lists" might actually be sets of multi-jurisdictional policies represented in RDF and/or OWL. So there is a direct tie-in to the semantic standards there.

Next is Hospital Availability (HAVE) which allows hospitals or other medical facilities to represent the number of beds available, the status of the emergency room, the surgical services available etc.

In an emergency, routing people to the right facility to support their need is crucial. The idea is that HAVE can help assist with that effort by ensuring the status of nearby medical facilities is up-to-date and accessible for machine-processing.

Next is Resource Messaging, a protocol for message exchange useful for requesting,, allocating and tracking resources, such as ambulances, fire trucks, etc. during an incident.

A standard of relevance in the works is Situation Reporting, which can be used by any organizational unit to report their current situation.

All of these standards are important for the domain of emergency management and significant amounts and degrees of decision-making are occurring throught the process.

Our work can assist these efforts by representing this decision-making, where the standards mentioned represent various aspects of the decision process, such as information inputs, decision result outputs, or metrics or options.

That's it for the quick overview of the EDXL standards supporting the emergency management domain.

To quickly finish our agenda,

Application of SPARQL for Decision Metric Assessment

SPARQL has significant flexibility in the filtering and ordering components, including boolean logic, regular expressions, and basic mathematical processing.

This flexibility means that SPARQL queries could support the assessment of decision options based on a set of criteria. We should consider that our decision representation format (including weighted criteria with options)

could be converted into a SPARQL query to perform a rank ordering of the options for an automatic assessment which can then be adjusted for a final selection by the decision-maker.

In this way, we come full circle, from defining the needed decision and the basis for making the decision and options (perhaps from browsing open data sources and picking the components to be used for those purposes)

to accessing the RDF data stores and applying the SPARQL query to assess the options, back to the final decision and representation of that final component.

In this way, our decision work not only leverages the open data work but also provides an important purpose for that work, to facilitate decision-making.

Thanks

Meeting adjourned.

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2010/05/13 15:45:54 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
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
Inferring ScribeNick: jeff
Found ScribeNick: jeff

WARNING: No "Present: ... " found!
Possibly Present: Eva Jeff Marion ScribeNick decision-xg eblomqvi left
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_4_Agenda
Got date from IRC log name: 13 May 2010
Guessing minutes URL: http://www.w3.org/2010/05/13-decision-xg-minutes.html
People with action items: 

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.


[End of scribe.perl diagnostic output]