decision-xg -- 27 May 2010

<scribe> scribe: jeffw

agend+ Open Linked Data Use Case: Progress and Next Steps

Agenda & Status

Over our initial meetings we've actually accomplished quite a bit. We've documented on our wiki and we've educated ourselves on the eXtreme Design (XD) approach to knowledge engineering, we've downloaded and explored the Neon Toolkit with the XD and SPARQL plugin, we've begun to expand on some initial use cases following the XD approach.

We've outlined on the wiki some key decision components and we have a sample underway where we used the tools to specialize the Transition pattern to address one of the use cases (information flow). We've also prepared wiki pages providing basic introductions to some of the semantic standards and related domain standards for emergency management like EDXL.

So where we are now is that we're ready to really begin to put some meat on the bones. We're ready to start a significant modeling effort for the 8 or 10 decision components we've identified. So this brings us to today's Agenda. I suggested the following topics:

Review of Status & Goals

eva: we will discover what is missing from use cases when we start modeling, we'll discover point

don: we have some basic use cases, so let's just dig some details

connection between use cases and components, shouldn't there be that?

eva: yes, we'll discover a missing use case as we go, for example recording decisions or core decision component, the idea is that each use case produces one or a couple small modules that tend to solve that particular problem

jeff summarized use cases

eva: I believe we are covering core decision representation and interoperability implying core representation of decisions and we were going to talk about event as a notion important for decision, so maybe they are in the use case there, or maybe
... I suggest 4 or 5 domain centered use cases, which result in 3 or 4 modules (interconnected ontologies), we don't need one use case for each individual small thing
... of course, we want to ensure we cover what we want to do with the decision format, because from those we derive the competency questions. And we can do this iteratively

don: I agree with eva, that we should just not necessarily add alot of stuff, but just break down the one

what do you recommend eva for how to do modeling, I'm not sure we all have time, but why not divide between us these

eva: important on the next telecon that we could discuss what patterns you want to reuse, these could be one way to synchronize between us that we are talking about the same things, that we have the same model in our head
... I propose we start modeling, and then we all present what patterns to reuse and if no patterns, then what are the modeling issues, and then we can post our draft solutions on the wiki
... a good idea is to post a picture of the solutions, so I need to check for visualization plug=in, cause a nice pictorial demo of the model is more intuitive than the OWL file, which we should also pass around but for a meeting better to have graphical illustration

don: I was thinking I could do situational awareness for emergency services

eva: I would prefer the open-linked data use case

use of transition pattern

jeff summarized his transition pattern work (see info at end of these notes)

example core component, event

jeff summarized his perspective on events (see info at end of these notes)

eva: I'm familiar with the concept and I could see what is done in the ontology domains, but heavier ontologies are trying to characterize events, but I could consider what is there and on the lightweight issue, one of the common misconceptions of ontologies
... and vocabularies on the web is that you need to use the complete thing, say you model events with time, place, etc. the whole idea with owl ontologies is that when you add data, you add only what is applicable
... even if restriction says an event always has atime, it doesn't matter if you don't add that time, you already have the possibility of an arbitrary sparse graph of data
... You can have optional things, so that you don't have to use the whole thing, in total it still is small and clear enough so that it is easily understandable

don: I think that for "event", we need to think more in terms of how managed program languages handle representation of events v. cursor on target and ucore, cause for cursor on target (what, where, when with option for details)
... I don't think it would be sufficient for this and if you add in more stuff, then you don't need it, so we should look at what is there and tie-in event based representation in computing

What would be an example?

don: Say you have a socket listener, one way is a thread to poll whether you have any data, another I have a general thing called an event and when you get data in this thing, come poke me
... So looking at the architecture for a generic construct but passing specific info into it would be good

I always like to have uris so you can include the info or just link to it

eva: if you have different components, so that you modularize your data as well, it's like the linked data, you can relate the dataset, but not always using the complete cloud of data, a similar structure here would be a good solution
... for next meeting, Don had some ideas of what we could look at, so why don't we do some collection of background material and we can model events with whatever until then. I'd like to look at existing event models and Don could put something on wiki about event driven software approach.

open linked data next steps

taxonomies and understanding them

eva: I will post something on the wiki about their different vocabularies

I also think we should continue to reachout to RPI and other folks working in this open linked data area as well as other W3C members

Where I think our work is important for the open linked data initiative is both that we can help expand the data sources that are rdf-enabled through our application of our work to domains like emergency management

and also we can provide an important use case for the open linked data, namely for making decisions.

That's all the time we have for today. Thanks to everyone for calling in.

(Here are some notes on topics above, since I couldn't scribe well while talking)

use cases

We have a set of use cases and the question is whether these are sufficient to begin and also whether they are structured at the right level. First, we have the Informaion Flow use case

which utilizes the "decision state" component of our format. The idea is that one moves through different states in the decision process and one should be able to represent that and query when a certain state

began or ended and determine how much time was spent in given states. Since our organizations want to improve information flow, it's important to track the amount of time needed to get and pass the information

and potentially develop a metric to measure whether our efforts to improve information flow are working. The Open Linked Data use case is intended to ensure and demonstrate that our decision format

can operate on and utilize linked data in its components, including as subject matter of decisions, as options, as criteria and the ability to make decisions across distributed sets of data utilizing the semantic standards

including SPARQL as a technique to drive the application of the criteria on the options to drive an assessment (an ordered list of the options based on the criteria).

The Interoperability use case is intended to show that decisions and their subject matter can be shared across tool sets utilizing these standards to enable interoperability.

The use case for Capturing Expertise is to suggest that if decisions are captured and managed in a standard format then anyone who is a good decision-maker, their decision components, such as criteria for making

a certain type of decision can be linked to and reused by others. In this way, expertise can be captured and reused.

The use case for Templates for Efficiency is to show that the same reuse of components generated by others can improve efficiency of the decision-making process as a whole.

Finally, the Situational Awareness use case is one where we recognize that decisions are being made constantly in our organizations and they are not well documented and managed, but they could be

and if they were, including the state of the decision-making, and aspects of decision-making such as which options are being contemplated currently, then situational awareness would be vastly improved and enabled.

(After some discussion that these use cases are sufficient to begin modeling, the participants picked their favorite to begin.)

<scribe> ACTION: Jeff will move forward to model the Information Flow use case. [recorded in http://www.w3.org/2010/05/27-decision-xg-minutes.html#action01]

<scribe> ACTION: Don will move forward to model the Situational Awareness use case and will report next time on what he learns about how managed program languages handle events in terms of processing/updating and the impact on the model. [recorded in http://www.w3.org/2010/05/27-decision-xg-minutes.html#action02]

<scribe> ACTION: Eva will move forward with the Open Linked Data use case and report next time on what she learns about taxonomies in use on this data. [recorded in http://www.w3.org/2010/05/27-decision-xg-minutes.html#action03]

Transition Pattern work

I used the Neon toolkit with the eXtreme Design (XD) plugin and the SPARQL plugin. I specialized the Transition pattern by importing it and subclassing a few of the classes to give them more specific names

for this application, e.g. DecisionState. I then did further subclassing to suggest that there might be sets of states from different perspectives and one of the perspectives is "Information Flow". Under that

subclass, I then added subclasses for types of states, such as InformationGathering or InformationAnalysis. I then moved forward to creat individuals/instances of those classes. So I created specific

time intervals, which I then linked in to specific instances of states for a given sample decision, and then I linked the states into some transitions. I then moved forward to use a SPARQL

query to show that I had represented the information that I needed in the form I needed it. This was a good first step for me and I learned much about the use of the tool. I will pair-design with

Eva as I move forward with this use case, since she has been kind to help me with several questions I've had regarding use of the toolkit, the plugin and the ontology patterns.

Open Linked Data Use Case

Two items worth mentioning: first, we've begun a reachout effort to some of the folks at RPI and we should continue to ensure that we proceed with this use case as well informed as possible and

that our thinking aligns with the work ongoing in this area. If we can get someone to speak at one of our meetings on this topic, if we can get someone who can be a resource for us to check

with as we proceed, and better yet, if we can get someone to participate in our incubator from this work, then that would be a great benefit hopefully to all participants.

Second, the taxonomies are key to utilizing the open data sets, in other words, what terms are used in the data sets. Our education on this topic will be key for us successfully tying into the data sets.

Another point is that there are tools and documentation on the RPI site about how to rdf-enable data sets, and this may be something we should do as part of our emergency management domain.

Meeting was adjourned on time and the final notes were added to flesh out the scribe's comments which were not captured at the time.

- DRAFT -

decision-xg

27 May 2010

Attendees

Contents