Ontolog Forum
| Number | 62 |
|---|---|
| Duration | 1 hour |
| Date/Time | January 7 2019 18:30 GMT |
| 10:30 PST/1:30pm EST | |
| 6:30pm GMT/7:30pm CET | |
| Convener | Mike Bennett |
IAOA Semantic Web Applied Ontology (SWAO) SIG
Meetings are normally on the first Monday of the month at these times.
- We will use GoToMeeting for this meeting - details at Connection Details
Agenda
1. Housekeeping
2. Events
- Proposed 'Challenge' activity for Ontology Summit 2019
- Financial Ontology Workshop in Summer 2019
- Shared events calendar - 2019 events
3. Discussion of new initiatives
- Bridging the gap between AO and SemWeb
- Definitions of ontology and related terms – IAOA glossary
- Blogging
- Networking
- Liaison with other bodies (IOF etc.)
4. AOB
5. Next Meeting
Proceedings
MikeBennett: Challenge activity for Summit
MikeBennett: The Sumit starts on Wed 16th
MikeBennett: Would help to be able to announce this Challenge if we have something.
MikeBennett: See 14:34 in last month's notes
MikeBennett: Challenge idea is based on the ESWC Requests for Challenges. KB sent them a request for advice (as they said you could do) but got no response.
MikeBennett: THis was not specifically a proposal for a challenge to do at ESWC.
MikeBennett: Challenge (sic) i to come up with such a proposal. Evaluation methodologies are not available.
MikeBennett: KB was proposing to submit a Challenge proposal to ESWC, rather than suggesting that the Summit itself runs a Challenge. Would write up something for the k/o meeting for the group to come up with some methodology for evaluating the quality of explanations for explainable AI.
MikeBennett: There was a talk at one of the summit sessions on 3 levels of explanation:
MikeBennett: 1. Technical explanation - describing the mathematical specification
MikeBennett: 2. gets progressively more into communicating with people (who are not familiar with the mathematics behind the ML algorithms).
MikeBennett: Good tech explanation is dependent on ontologies. Enables a narratiev of dialog with the individual in explaining what is happening.
MikeBennett: Presenter works in an XAI program. People do not have a clear what an explanation is, or don't have the same idea of what is required for a solution.
MikeBennett: Hence needed to reiterate the notion of evaluation criteria. Other than simply asking if a person is happy with the explanation. Else the state of the art for explanation for AI systems is not well developed.
MikeBennett: Given ISWC deadline of 15 Jan, we would not have the chance for someone in the Summit to work together on submitting a proposal.
MikeBennett: The issue is that there are no evaluation criteria.
MikeBennett: ESWC was Dec 15 deasline for proposals.
MikeBennett: ISWC is 15 Jan
MikeBennett: If ISWC accepted our proposal then once we know f it accepted we could encourage folks at the Summit to submit a response.
MikeBennett: We can mention that it is proposed and what the submission date for entries is.
MikeBennett: If not accepted for ESWC or ISWC it can be done at the Summit anyway.
MikeBennett: Ken will take the proposal and add it to the agenda for the 16 Jan k/o meeting.
MikeBennett: KB submitted something to ESWC via EasyChair rather than as an email. Needs some feedback.
MikeBennett: KB We can describe the Organizing Committee of the Summit to be the Org Ctee for the Challenge. This is already in the draft proposal.
MikeBennett: So there may be enough already to meet the requirements of a Challenge proposal to one of the other. If we get any feedback from ESWC we can include it in the submission for ISWC.
MikeBennett: A weakness may be that the submissions require some evaluation criteria but this is the very thing no-one has.
MikeBennett: If we had the meta-criteria that could be framed as a challenge in its own right. But we do not.
MikeBennett: Or we can run the meta-challenge within the Summit itself as a challenge.
MikeBennett: KB can mention this on the 16 Jan call.
MikeBennett: IF we had some clue as to how we would evaluate an explanation this would help indicate how an explanatiojn is constructed.
MikeBennett: For example there is a knowledge aspect, a social aspect, linguistic aspects. Mathematical aspect (see above)
MikeBennett: This can be done at the Summit as a workshop, or as one of the workshop problems for the face to face.
KenBaclawski: The Doron session of the summit is relevant: http://ontologforum.org/index.php/ConferenceCall_2018_11_28
ToddSchneider: https://www.dnvgl.com/events/international-industrial-ontologies-workshops-134292
MikeBennett: This is the IOF event in Oslo in 1st week of Feb.
MikeBennett: Oslo, Norway on 4 - 7 February 2019
MikeBennett: Barry Smith pushing IOF to adopt BFO as their universal ontology for this. Some push-back. BS expected a decision on this in November. Some discussion will follow. MG pushing to get some requirements for the proposed industrial ontology.
MikeBennett: BFO has no Change Request mechanism. Not clear how that aligns with the notion of an ISO standard. Proposal seems to be for extensions and as when required.
MikeBennett: Need to drop the Realism requirement and allow for abstract notion.
MikeBennett: Should we as a group be looking to do something on the ISO TLO standard liaison, feedback etc.?
MikeBennett: Given we are also interfacing with Semantic Web, can we do something on TLO for that community?
MikeBennett: MB Yes; some SemWeb folks run scared of TLO perhaps without good reason e.g. misaapplication of these concepts at client sites.
MikeBennett: TS: WE can come up with examples of good usage and describe the representation challenges. e.g. what a function is, when realized. how to treat in BFO (concretized) v others.
MikeBennett: This is explaining an UO which is a kind of explaining something. How we can you explain an AO and does it contribute to the explanation. MAy be a contributiojn to understanding what UO items to use.
MikeBennett: What are the criteria for having a good explanation? Just giving an abstract account of the UO elements may not be practical.
MikeBennett: So there is explainable UO versus Explainable AI.
MikeBennett: BFO 2.0 - one important aspect is documentation. Many TLOs fall short on that.
MikeBennett: Can we create a resource on the use, common mistakes etc. for TLO elements.
MikeBennett: All agree this is a good idea.
MikeBennett: NExt steps:
MikeBennett: Collect some, review them, think about how to present these.
MikeBennett: Put a bunch of things into a review of a paticular example/
MikeBennett: MB: Identify the common problems, e.g. behind Continuant v Occurrent; relatiev things et. and describe how the leading ontologies address these common issues in different ways.
MikeBennett: What venue?
MikeBennett: Something on the IAOA website.
MikeBennett: A blog?
MikeBennett: Something where people can put comments and responses. A blog (as configured in our wiki structure) would cover this.
MikeBennett: WE can do this within our SWAO Wiki structure and promote it to the IAOA once it is mature enough.
MikeBennett: Review when or whether to allow for a response capability - given it can get heated and the heatdedness can turn other participant off. Hence not a mailer.
MikeBennett: First collect some material.
MikeBennett: Topics:
MikeBennett: Things that happen (Continuant v Occurrent);
MikeBennett: Thins in contexts (Relative Things / qua entities)
MikeBennett: Units and measures (Sweet ontology etc.)
MikeBennett: Make clear we are not taking a 4D approach, not treating everything as a process - i.e. not being locked in to one or another practitioner's set of assumptionsm OR when we are, make a given ssuption or decision explicit. Help people understand the decisions and modeling choices.
MikeBennett: e.g. in IOF, the different people have different notions even of wht is ontological analysis versus modeling design decisions.
MikeBennett: We need to present the options, assumptions etc. rather than define a right and wrong approach to these things.
MikeBennett: List what the TLO considerations are as above;
MikeBennett: List what are the paradigms within which one has to think these things: 4D v 3D; Realist or not; treatments fo time and temporality and so on. Whether you can have multiple inheritance
MikeBennett: Action Circulate the list of problem areas and the list of pardigmatic considerations, as outlined above, via email, then start to put a wiki page together.
MikeBennett: AoB?
MikeBennett: Next Meeting?
MikeBennett: 4 Feb (during the IOF thing in OSLO)
MikeBennett: Defer to 11 Feb?
MikeBennett: Separately, Andrea can't make this time slot generally.
MikeBennett: We will move these calls to 2 - 3 going forward. Create a new / update existing Outlook entry.
MikeBennett: Add this to the Ontology Events Google Calendar.
MikeBennett: This should go in the IAOA Google Calendar. Can also go in the Shed and SWAO spreadsheet-generated one.
MikeBennett: MB will update the 2019 Events spreadsheet and share with IAOA. They may choose to include or ignore some of the more specialist stuff.
MikeBennett: Can include AI, Blockchain, other things of relevance to ontologists but not directly about ontology.
MikeBennett: Confirmed we will meet on 11th.