| Number | 79 |
|---|---|
| Duration | 1 hour60 minute 3,600 second 0.0417 day |
| Date/Time | June 8 2020 18:00 GMT |
| 11:00 PDT/2:00pm EDT | |
| 7:00pm BST/8:00pm CEST | |
| Convener | Mike Bennett |
IAOA Semantic Web Applied Ontology (SWAO) SIG
NOTICE: This month's call has been rescheduled to Monday 8 June (next week, same time) due to a conflict.
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
- Status and updates
- The IAOA EC has asked that we post a brief summary on the IAOA website each month.
- Current progress and prospects for the special issue on explanation.
- Housekeeping
- The soaphub site is no longer reliably available. What is the long term solution?
- AoB
- Next Meeting
Proceedings
Status and Updates:
IAOA EC Request
Clarify: which EC? Executive Committee
Note that Education Committee is now abbreviated EDC
(Note that EDC has been a bit quiet).
Next EC meeting tomorrow 9 June.
They would like us to post (put up a formal posting) once a month of a summary of SWAO activities.
Q: Would meeting minutes suffice for that?
A: No.
They want to show activities going on.
Also this would be a mechanism for recruiting new members.
Attracting New Members
We need to do something to attract new members; we need new members to do anything.
So we should have something to show, that requires only existing members, and can attract new ones.
Special Edition
KB: We can present the Special Edition activity as such an activity.
Q: Up to now that has not been an activity of this SIG. Are we now appropriating this?
Membership differs - Ram sri Ram and Gary Berg Cross are not members / regular attendees of this SIG.
How would we sell this as a proposition to new members of this SIG?
We would have to do a lot of spinning to make this work.
What would we be asking new people to do?
This post is only to show what we have done, not something we are asking new members to help with.
Resolution: We can say that this SIG is 'participating in' the Special Issue.
SWAO SIG Purpose
Our real purpose is not only the Semantic Web but other groups too
Historically included Big Data.
Potentially also AI and so on.
TS: Also software development community - so the connection with the Semantic Web is not distant.
“Forum for multiple communities to work collaboratively in tackling common problems.“
(from the mission statement)
Examples:
- Semantic Web
- Big Data
- Knowledge Graph
- Linked Data
This is all on the general description page.
Charter is separate.
Written before Knowledge Graph became a buzzword.
So the Special Issue does fit into this framework.
MB in the things we have done, apart from our earlier Special Issue, we have never really originated an activity at this SIG, we've really worked on helping with other thinigs, housekeeping and so on over the years.
We need something we are about to do that people would want to join in order to participate in.
Ontology Analysis as new SWAO SIG Focus
TS Proposal: What if we spearhead this notion of analysis (i.e. how to figure out the semantics you need to represent, as distinct from just OWLifying vocabulary).
TS has processes for analysing natural language with a goal of developing ontology. Has a linguistics art and an ontological partr. This is in progress. Slanted to IOF so geared towards use of BFO.
MB also saw a good thing from Barry Smith on ontology development guidelines, but again this is also somewhat specific to the BFO 'REalist' approach.
See also Stefan Schultz and Ludger Janssen and others. TS did a review of this a while back.
MB: IT would be good to have a more generalised set of ontology development guidelines that are not locked into BFO's type of Realism.
TS: Also there's the challenge in using natural language terms as the signature for a concept, which makes it hard to divorce that understanding from the symbol.
Example Topic: Word Abuse
KB: The notion of 'Word abuse' given in a talk by Robert Hoffman at Stanford CS520 recently.
KB has been objecting to this kind of thing for decades.
TS asks to clarify.
JB: Example: 'What's the connection between an IQ test and intelligence'. Answer was 'intelligence is defined to be your score on an IQ test'. Unexpected answer.
Other research indicates low correlation between the natural definition of intelligence and IQ scores.
People end up getting grants (pragmatics) for IQ on the false premise that they are looking at intelligence as generally understood.
Also the word 'intelligence' in AI definitely does not map to IQ.
MB there's almost nothing about an AI system that maps to IQ.
TS: Examples of this in BFO v2.
What was intended is often completely different to what was expected.
BFO overcomes this by having non meaningful labels.
MB: in ontology all we can do is define concepts and give them words as labels. We can't define the 'actual' meaning of a word across the whole semantic space that that word maps to.
Ontology Analysis
So the above side conversation on ‘word abuse’ was today's activity.
What shall we call this activity as seen above?
KB: There are a number of sources for the things above, e.g. an article by Todd.
What about if we got MB and TS to present their views on the above and present to the group. That might give us a focus in completing some of the thoughts of BM and TS, i.e. an incentive to complete the documentation of this thinking
KB: Existing development methodologies seem to focus on the syntax not the semantics.
MB: BS's paper is a good example of the right focus, though constrained to a given paradigm.
ToddSchneider: 'Guideline on Developing Good Ontologies in the Biomedical Domain with Description Logics'; URL: http://www.purl.org/goodod/guideline
KB: Good example in Ontology101 'Burgundy is a kind of wine'
We need heuristics and guidelines. Collect and disseminate these so that all the communities doing semantics we do it the same way.
MB this would be a huge contribution to the industry.
Smith has covered some of this but not all (nothing about restrictions etc.).
Conclusions
So we have an activity.
Need a name for it.
This is our new project.
We had our first discussion on it today. This is an ongoing activity.
Examples:
Monohierarchy v polyhierarchy
What to call it examples: Principles, Guidelines, Heuristics...
Collected wisdom.
e.g. principle of monohierarchy. Pros and cons.
Get away from the doctrinaire approach where someone thinks their own usage is the only correct one.
ToddSchneider: "Best Practices of Ontology Development", available from https://drive.google.com/file/d/1xV8KNp2mpruQF2s07-WcyYE5uux64QVW/view
KB: Smith talks about Normalization, to deal with combination classes. Have first to normalize your hierarchy into monohierarchies and then have rules for how to combine concepts from the different monohierarchies, leading to a polyhierarchy but you can see from each term where it came from.
MB has written extensively about mon v poly but not about normalization in the sense given in Smith. So we should look at that.
TS: Don't discount the value of BFO and others.
MB strongly agree - we should be able to have guidelines that work across and between the different ontological stances, for example knowing when to use realism.
TS: Also when to use a particular TLO in your analysis.
Also other methodological matters e.g. hub and spoke (now Snowflake).
KB: How you extend the snowflake. Its hierarchy would be a tree with a root node attached to one of the leaf nodes of your snowflake.
TS: That's nonsense.
That's an example of a certain technique that is useful in certain cases.
KB research of category theory for KGs. When you do the above and extend from the leaf node, you get a pushful transformation
A pushful transformation causes changes at other levels.
Multi-level hierarchies.
in the sense of meta-levels not in the sense of abstractions
Clarify do we mean meta-levels or abstraction levels?
We mean meta-levels here.
So you end up inducing transformations on lower levels. Doesn't happen just for any kind of design but only for the hub and spoke design.
Unexpected advantage.
Guidelines - semantic.
Come up with a workstream.
We can report back on that tomorrow.
Name for this new Activity: Needs to include 'analysis' so people get away from the idea that it is a cookie cutter approach.
KB likes 'guideline'
'Semantic Analysis Guidelines'?
KB: Analysis too many-meaninged
Ontological modeling guidelines?
Pre-cursor to modeling.
Principles.
That would replace 'guidelines' not 'modeling' or 'analysis' which is the word we are trying to deal with.
Example of 'ontology path dependence principle' which is a Principle.
Guidelines or Principles - not a problem
The problem word is Analysis.
Ontology Development Guidelines.
Ontology articulation guidelines
Analysis - was too reactive, need something proactive.
Hence Development, or Articulation or something.
Outcome: 'Ontology Articulation Guidelines'.
Housekeeping
SoapHub Unknown, so best not to rely on this.
For these meetings, we will continue to use GoToMeeting.
Next Meeting
July 6
(no holidays that day but is after the July 4 weekend)
Same time, July 6.