20170731 Ontology Change Improvement Call

Date: July 31, 2017

Attendees: Anna Kasprzik, Tatiana Walther, Marijane White, Mike Conlon, DJ Lee

Agenda:

1. Plans for meeting at the conference
2. Update on action items
3. Process for making changes – see Ontology Change Process (draft) https://goo.gl/3ZMSwC

Mike: Wanted to talk about possibly getting ontology people together at the conference. Christian will be at the conference. Marijane, you will be there? And Javed will be there, and DJ will be there, and Damaris will be there.

Tatiana: from the TIB, it will be only Christian. He will attend a Birds of a Feather meeting if there is one.

Mike: Yes, and so I wanted to think a bit about what might be useful to discuss when a number of us are going to be in one place at one time.

Let's go to the higher order question -- is it a good idea to meet at the conference?

Marijane: I think so.

Mike: I think so too. There are a number of technical issues that this group will hit almost immediately.
Let's meet for lunch on Thursday. Can send out a notice.

Some updates on the action items. So on the agenda, we actually are making progress. We need to create a viable ISF master that has all the VIVO ontologies. I have some slides to share, been working with ROBOT and have a good handle on what needs to be done. We have some reviewers and committers, that's good. Have a folder of all the ontologies from 1.92 that people can look at. Have been moving issues around between repositories. 18 issues in the community repo, 47 in the ISF repo, and 45 in the JIRA.

Marijane: I noticed the activity. Did you leave the ones that other people opened?

Mike: no, I moved all of them, including comments, made notes about who made original issue/comment. Tagged them all VIVO, except ones that seemed to be applicable to the entire ISF, and a couple tagged with "qustion" because that's what they are.

Javed is going to present his ontology for ontology changes in a future meeting -- has he shared his paper?

Marijane: no, i don't think he has. I don't see a link in my notes from our last meeting, anyway.

Mike: ROBOT tool -- I've been working with it, and I think everyone should try it. It finds things you can't find in Protege. It will help us build better ontologies.

Anna: Can you suggest some test data?

Mike: Yes, I recommend running the examples (as Marijane suggested to me) and then start running them on VIVO files.

Anna: Can it do diffs on different formats?

Mike: Yes. it does ontological diffs, independent of file format.

Anna: and it's not dependent on order?

Mike: No. It is working a graph.

It's concept of extract is interesting, different from what we might have expected. It has a very ontological view.

Marijane: I am not sure what you mean by ontological view, but I know it does a sort of MIREOT top/bottom extract.

Mike: it has several different extract methods, four i think? What I mean by ontological -- it wants to do something like an ontological closure. So if you say, "get my foaf:Person", well in our ontology foaf:Person has assertions about vcards. So it will make sure you have enough to represent the concepts you're representing.

We should definitely talk about ROBOT and proceed carefully with it.

And then Anna had a question about roles for committers/reviewers, we should also talk about. that.

Let me share my screen, we'll talk about the issue of having a viable master.

https://docs.google.com/presentation/d/1Zu7pylOzb-ZmEeJXqonn5ZLZeuu8Vn-d5pSylID_h1M/edit#slide=id.g1fa528f15f_0_6

Was interested in exploring difference between what VIVO has and openRIF has. Turned out VIVO has several sources and it wasn't clear which one we should be using. We have a file on the web at vivoweb.org/ontology/core, which is the file used in linked data requests. Supposed to be an authoritative source of the VIVO ontology.

The export has more axioms because it has inferences and Vitro classes. But it has the same number of classes, that's good.

The next one is the filegraph, which is the set of files we distribute as our ontology before they are loaded into a VIVO. They do not have inferences. Slight difference in class count, obj prop, data prop, and it has the individuals.

Javed has an analysis that is basically the same as the web file.

VIVO-ISF has a lot more axioms and classes, and interestingly not many properties. suspiciously few.

Then merged VIVO-ISF with the filegraph, just for fun and because that's what we distribute and gets loaded into VIVO. It added almost 1000 axioms, a bunch of classes, a bunch of data properties, and no individuals.

And then I diffed the source and the merge to discover what's going on. VIVO-ISF has 339 annotation assertion differences. They're all eagle-i. People creating better labels for things we use in common, perhaps shouldn't have said eagle-i. Better annotations. Filegraph has 19 annotation properties that are not in ISF. 67 classes, 8 that are BFO, that are not used in the ISF. 12 are VIVO and 47 are vcard. 11 data properties, 4 datatypes, 47 object properties, and 1000+ other axioms.

The 12 classes are mostly people classes, but but also core:Abstract and core:OrganizaingProcess and the advisee/advisor roles. Haven't explored source code to see if we're actually using OrganizingProcess.

So that's ROBOT. It will help us tremendously. Had to patch some things in the process, which will lead to us fixing things that need to fix, such as a small thing with respect to ORCiDs, some parts of the filegraph can't stand alone, ROBOT will assume some properties are annotations if they don't have object property annotations. We had terms but we hadn't said what they were.

I know Javed is very big on the idea of having organized subsets, oh look there's Javed now.

pause to let Javed connect audio

So Javed, I just finished a presentation talking about ROBOT and how I used it to compare various sources of the VIVO ontology. The takeaway is that there are more differences than we might have thought, but with ROBOT we can find them and fix them, and I recommend that you try ROBOT out.

Note that the slide with the missing classes is not the only thing that is missing, there are axioms and such that are also missing. But we have the tools to fix it, and

We also thought it would be useful to meet at the conference

Marijane: I might ask you for a little in-person demo of the stuff you did with ROBOT.

Mike: absolutely, I can walk you right through it. Can talk about Merge and Diff and Extract.

Oh, ROBOT also has a pretty good annotation tool that might be good for providing internationalization of labels in VIVO. It's very easy to get a list of all the classes and identify where you need labels.

Marijane: Could we also review the data that Dave Eichmann sent us this morning?

Mike: Yes, that's a great idea.

Marijane: For those of you not on the email to me and Mike, Dave has scraped VIVOs and crunched the ontology usage numbers.

Mike: Dave has scraped 30 different sites and computed ontology coverage. The neat thing is that it includes local extensions.

Javed: so Mike, your slides, can you explain the first page?

Mike: web is the file on the web, export is what you get when you export ontologies from a running 1.9.2 VIVO, filegraph is what gets loaded into a vivo, vivo-isf is the source.owl file from the repository, and the last one is a merger of the filegraph and vivo-isf.

Javed: Did you use a tool for this?

Mike: Yes, this is the result of a bunch of work with ROBOT. describes steps used to conduct analysis

Javed: When I've merged I've noticed that Protege creates new axioms, wanted to make sure that when you merged them that no new axioms were created.

Mike: ROBOT can detect all kinds of things.

Javed: A demo at the BoF lunch would be great.

Mike: I put Dave's files up at the Google Drive folder.

• Mike shared his screen and we looked at Dave's data *

Lots of great data! We're going to get really great control over the ontology.

Time to wrap up! If you haven't made comments on the process, please do so, and see you at the conference this week!