Session	Advantages of Formal Definitions
Duration	1.5 hour
Date/Time	17 Mar 2021 16:00 GMT
	9:00am PDT/12:00pm EDT
	4:00pm GMT/5:00pm CET
Convener	Alex Shkotin
Track	B

Ontology Summit 2021 Advantages of Formal Definitions

Ontologies are a rich and versatile well documented formal construct. They can be processed with algorithms programmatically. This session will explore the many kinds of ontologies and how they can be formally manipulated. The goal is to acquaint both current and potential users of ontologies with the possibilities for how formal processes could be used for solving problems. In ontology development and application, formal definitions enable precision and computer readability. We explore the advantages of formal definitions and make the case for why they are necessary in the ontology domain.

Agenda

12:00 - 12:20 EDT Selja Seppälä. Systematizing definitions in ontologies. Slides
- This presentation will outline a practical proposal to systematize and harmonize definitions in ontologies. The methodology can be applied both to textual and logical definitions to accelerate their creation. It may also be used for definition checking, quality control of ontologies, and automatic generation of definitions and axioms.
- Selja Seppälä, PhD, is a Marie Skłodowska-Curie Career-FIT Fellow at University College Cork, Ireland. She holds a PhD in Multilingual Information Processing from the University of Geneva, Switzerland. Her interdisciplinary research in the areas of natural language processing (NLP), terminology, and applied ontology focuses on definitions and lays the groundwork for creating computer-assisted natural language definition writing and checking tools leveraging ontological data. She has co-authored a number of publications on definitions in ontologies, and initiated and co-organized the IWOOD workshop series on Definitions in Ontologies (2013-2016).
12:20 - 12:40 EDT Andrea Westerinen. Experiences with Foundational, Domain and Application Ontologies in OWL2. Slides
- Ontologies are formal definitions of a domain of interest. They include the fundamental concepts, relationships and truths (axioms) that are relevant in the domain, as determined by the use cases that the ontology is meant to address. To specify an ontology, many aspects should be considered that strengthen its relevance for users, and domain experts, as well as its ability to support inference, computation and even argumentation. This presentation touches on many of these aspects based on several examples.
- Andrea Westerinen is a software and knowledge engineer, and CTO of OntoInsights, LLC. Her research focuses on combining semantic and natural language technologies with machine learning. Ms. Westerinen has worked in the computer industry since 1979, starting at the Jet Propulsion Laboratory, and including positions at Raytheon BBN, SAIC, CA Technologies, Microsoft, Cisco and IBM. Over the years, she has published numerous articles and been active in many standards organizations. Ms. Westerinen holds a Bachelor of Science degree from Marquette University and a Masters in Computer Science from Nova Southeastern University.
12:40 - 13:00 EDT Caitlin Woods, Emily Low. Representing ontologies in Natural Language and First-Order Logic: Why do we need both? Slides in pptx format
- This presentation explores how natural language (or “Subject Matter Expert”) ontological definitions can (and should) be translated into First-Order Logic (FOL) as part of the ontology development process. We use examples from the engineering domain to demonstrate the value of the clarity provided by the FOL definition and some common translation mistakes. We also discuss how translation can, and should, be treated as a collaborative process. Finally, we share some of the lessons learned when using this translation process in our own ontology development project.
- Caitlin Woods (software engineer and PhD candidate) and Emily Low (mathematics and philosophy honours graduate) are from The University of Western Australia. They work in a research lab (The UWA System Health Lab) under the supervision of Professor Melinda Hodkiewicz and are active members of the Industrial Ontologies Foundry. In 2020, the System Health Lab started a blog called Ontology Explained. The purpose of this blog is to share the lessons that they have learned over the past few years in a way that is familiar to engineers.
13:00 - 13:30 EDT Discussion
Video Recording

Conference Call Information

Date: Wednesday, 17 Mar 2021
Start Time: 9:00am PDT / 12:00pm EDT / 5:00pm CET / 4:00pm GMT / 1600 UTC
- Ref: World Clock
- Note: The US and Canada are on Daylight Saving Time while Europe has not yet changed.
Expected Call Duration: 1.5 hours
The Video Conference URL is https://bit.ly/3i1uPRl
- iPhone one-tap :
  - +16465588656,,83077436914#,,,,,,0#,,822275# US (New York)
  - +13017158592,,83077436914#,,,,,,0#,,822275# US (Germantown)
- Telephone:
  - Find your local number: https://us02web.zoom.us/u/kd3SqdIOZl
  - Meeting ID: 830 7743 6914
  - Passcode: 822275
Chat Room: https://bit.ly/39PzQJW
- If the chat room is not available, then use the Zoom chat room.

Attendees

Discussion

[12:08] RaviSharma: Selja does that mean function is a definition?

[12:11] RaviSharma: so more than one of the function characteristics can be picked up?

[12:18] RaviSharma: example is fine for architecture implying some kind of facility may be a mobile van or a tent in a village or regular hospital facility. We get the point you are making namely there are drill don categories or classes that further narrows the type of function.

[12:21] RaviSharma: Yes very good for autogeneration from general templates and classes.

[12:22] RaviSharma: What is the role of semantics in such automation, how is semantic relevance measured?

[12:24] Joep van Genuchten: as a practitioner, i find that those who understand how to interpret ontologies, typically ignore the textual definitions, those who do not understand then (typically Subject Matter Experts) find natural language definitions much more intuitive (they typically don't understand axiom based text either). How do you look at the use of more natural language oriented definitions aimed at the intuition of SMEs?

[12:27] RaviSharma: Andrea - Does that mean that events and processes are to be handled differently than static ontologies?

[12:28] John Sowa: This talk verifies what I have been saying for years: BFO is worse than WordNet or just a typical dictionary.

[12:29] Mark Underwood: "The Sixties were cool." Heck yes.

[12:31] John Sowa: In fact, I would say the same for most top-level ontologies: they're useful only for a tiny minority of developers

[12:31] RaviSharma: Andrea - I have been saying that in these days A_V learning can perhaps beat the text based comprehensions and smartphones and TV etc and SMEs prove more in that direction and it also implies multiple language speaking people can communicate better and events and stories are relevant.

[12:32] RaviSharma: John - how does this relate to Ontology pyramid in your slides ESWC??

[12:34] John Sowa: Ravi, I would relate this to the final section about automated and semi-automated methods.

[12:34] John Sowa: The computer should generate the ontology automagically.

[12:35] John Sowa: But the human should be the guide in helping the computer system and answering any questions it has about intentions

[12:36] RaviSharma: John - this is the situation with events, but what about complex event processing and situational awareness semantics?

[12:39] RaviSharma: Andrea what you show on slide 9 is more relevant to static situation but dynamic situations and events should have influence on some of these that you list?

[12:40] RaviSharma: John- Yes fully agree.

[12:40] John Sowa: Ravi, The same kinds of problems arise with everything. See Section 7 of http://jfsowa.com/talks/eswc.pdf

[12:41] John Sowa: Notice that Andrea is talking about inconsistencies That is inevitable in EVERY human specified ontology

[12:43] RaviSharma: John - what is role of inferences and reasoning in semantics - is it meaning interpretation or knowing that which is important in the context or domain?

[12:43] John Sowa: It is much easier for a computer to generate an ontology is guaranteed (by construction) to be consistent

[12:45] RaviSharma: Andrea- great analysis

[12:47] Andrea Westerinen: @ravi Events and processes are defined in the ontology. The ontology gives you the concepts, relationships, core axioms and overall structure. I would say that events/processes have a definition.

[12:49] Andrea Westerinen: @john I try to create subclasses and instances using VerbNet, WordNet, etc. based on the high-level ontologies that were defined.

[12:49] Joep van Genuchten: Andrea: the geospatial has very elaborate and taxonomies for expressing shapes. If we look at time as a 4th dimension: what would 'spatial geometries' look like in terms of events and states?

[12:49] janet singer: @Mark : )

[12:50] Andrea Westerinen: @joep Events and states occur in spatial geometries. I have purposefully left the top level broad and open in order to replace location/time/agent/resource definitions with what is relevant to the domain.

[12:52] Joep van Genuchten: its, my late afternoon, and my brain is malfunctioning: The question was supposed to be: what would 'temporal geometries' look like in terms of events and states?

[12:52] Andrea Westerinen: @ravi Slide 9 is about event types but all narratives describe dynamic/different sets of occurrences.

[12:55] ToddSchneider: Caitlin, Isn't there an assumption that the natural language terms or phrases used in the logical definitions (i.e. the signature) will have a natural language 'understanding' that is sufficiently common among all the participants in the ontology development process?

[12:55] John Sowa: I agree with Caitlin that FOL is essential for communication among machine components. I strongly recommend the DOL software for mapping all Sem Web notations to and from FOL or CL.

[12:55] RaviSharma: Caitlin - slide 5 raised it but slide 6 answered it namely using FOL to understand data connections?

[12:55] Andrea Westerinen: @joep I might not be understanding your question. Here is an example ... A fire occurred at the DuPont factory on May 1 2019 ... the event is a kind of disaster (fire) that human reasoning knows causes damage. The location is a Building (the DuPont factory) which may have a known address, GPS location, ... There is a PointInTime which is a day.

[12:56] Andrea Westerinen: @joep I utilize Wikidata, etc. to add to the stated details in a narrative.

[12:57] Andrea Westerinen: It is totally amazing what is now available as open-source.

[12:58] Mark Underwood: @Andrea We (ontologists) may have to pay for it going forward https://www.wired.com/story/wikipedia-finally-asking-big-tech-to-pay-up/

[13:01] John Sowa: I like Caitlin's slide that shows three kinds of people: Software engineer, Mathematician/philosopher, and Subject matter expert

[13:03] John Sowa: But I would replace all three of them by a computer system that uses FOL internally and interacts with humans by a dialog in natural language

[13:03] RaviSharma: Caitlin - what is an example of Edge-case?

[13:04] Joep van Genuchten: @Caitlin: any insight on how to deal with aboutness in discussions with SMEs (who doesn't think in terms of descriptive content entities, but in terms of the things the description is about)?

[13:06] Andrea Westerinen: @mark Most APIs either have a usage cutoff or pay option. I try to live within the usage cutoff.

[13:07] Douglas R. Miles: @Andrea.. you are using a bit of event calculus?

[13:08] Douglas R. Miles: (Well i mean to say you described situations in which you would use)

[13:11] Douglas R. Miles: my question is if you have found any good tools to do such event calculus reasoning over narratives

[13:11] TerryLongstreth: Andrea: is evolution an event?

[13:13] Andrea Westerinen: @terry I subclass process to natural process (e.g., evolution) and goal-directed process (e.g., a business process). Evolution is a set of changes, but a change (or recognition of a change) is an event.

[13:16] Joep van Genuchten: @john: doesn't what you say about natural language based ontologies assume that natural languages and formal languages have similar expressivity, and isn't that a rather large assumption?

[13:16] ToddSchneider: Have to leave. Thank you.

[13:17] Andrea Westerinen: @douglas I am playing with several different packages but have not settled on one. There are also some very cool causal analysis routines.

[13:18] Andrea Westerinen: @douglas A lot of what I have been playing with, I plan to open-source ... including the Event-State ontologies.

[13:19] TerryLongstreth: @Andrea in natural language many 'processes' are often subsumed as events; evolution happens to a genus, refinement happens to crude oil,

[13:19] Joep van Genuchten: I got to go as well, thank you

[13:21] Andrea Westerinen: @terry I agree that natural language is ambiguous, but that does not mean that the ontology is

[13:21] Andrea Westerinen: For me, the ontology is under the covers ... to achieve narrative analysis. It is not the end goal.

[13:29] RaviSharma: thanks to speakers for their great inputs and efforts

[13:31] TerryLongstreth: @andrea, thanks for your insight. I wasn't trying to be snarky, just highlight that process and event are sometimes closely related or even alternative views of the same thing

[13:31] Andrea Westerinen: Thanks to all for the dialog.

[13:32] RaviSharma: Alex Shkotin - thanks for your efforts

[13:32] AlexShkotin: I am happy

[13:35] AlexShkotin: have to leave. Thank you all.

[13:37] TerryLongstreth: A better example of process vs. event is "heart-attack"

[13:38] Cas Miles: Selja, Andrea and Caitlin thank you very much- very interesting and informative!

[13:38] Andrea Westerinen: @terry So is heart-attack an event with component events?

[13:39] TerryLongstreth: ask a doctor, they'll probably say 'both'. He had a heart-attack. It started with chest-pains.

[13:40] TerryLongstreth: or maybe, it started with atherosclerotic blockage

Resources

Selja Seppälä. Systematizing definitions in ontologies. Slides
Andrea Westerinen. Experiences with Foundational, Domain and Application Ontologies in OWL2. Slides
Caitlin Woods, Emily Low. Representing ontologies in Natural Language and First-Order Logic: Why do we need both? Slides in pptx format
Video Recording

Previous Meetings

... further results

Next Meetings