20181018 Ontology Change Improvement Call

2018.10.18

Attendees: Brian Lowe, Ralph O'Flinn, Damaris Murry, Marijane White

We're reviewing the old Change Process proposal: https://docs.google.com/document/d/1FVxKmYHfDglgrBHseKcssBZ5NUXMgQ9lBJVWyNpgpdA/edit#

This has not been touched in over a year. At the time it was written, the VIVO development team was not as coherent as it is now, and as such it doesn't even address the possibility of bringing changes to the development team. Do we want to try to edit this, or do we need to start fresh with a new proposal?

Damaris notes that this came out of concerns about communicating the impact of changes that resulted out of the move to version 1.7 of the ontology, where many sites never upgraded because of lack of capacity to respond to the changes. We need a template for communicating changes in a standard and clear way.

There is concern about communication of ontology changes going through the developer team, and how well that might work.

Brian made some very good comments echoing and expanding on what Damaris said that Marijane mostly missed capturing for the notes.

Right now the product evolution team is very focused on ease of development, rather than the ontology.

Brian remarks that VIVO was initially intended to be a wholly semantic application. We need a robust ontology engineering effort, and if we can't do that then maybe we shouldn't do it at all.

Ralph suggests that there could be something added to improve ease of development work with the ontology. Ralph thinks the product evolution group, the development team, and the ontology group need to get together.

Marijane is curious about how the software interacts with the ontology. How direct is it, is there an application layer? Is Vitro that layer or is it just an editor?

Damaris' experience is that a lot of stuff cannot be changed in the UI because of the ontology.

Ralph considers VIVO as a layer on top of Vitro.

Brian says it's a bit of a hodgepodge. Stuff like SPARQL queries have been baked in over time. Some things are configurable, some are not. Some are configurable in theory. Overall, there are functionalities in both VIVO and Vitro, but there are no abstractions of say, a person, so developers are really interacting directly with the RDF, lots is done with SPARQL queries. An abstraction layer might be a good thing to add in the future, but we don't want to make the mistake of mixing those concerns as may have happened in the past.

Ralph notes the new data distribution API and other newer pieces that are mentioned in the wiki but may need better explanations. A lot of stuff is discussed in these calls that doesn't get captured elsewhere.

Damaris feels like the ontology group is at the heart of what VIVO was intended to be. The Product Evolution group has good intentions but they don't seem to know how to work with this group. This group has a responsibility to communicate what the ontology is for and why it's important, the PE group needs our guidance.

Ralph says the PE team has focused on what data they want to see in their proof of concept, which stopped discussions of what technologies to use, and they started working on a data model, how it should be structured, and what technologies should be used to use that model in a dev friendly environment. They are in the middle of steps one and two.

Marijane heard a rumor that the PE group was considering moving away from ontology. Ralph says they're not quite doing that, they are trying to remain faithful to the ontology while coming up with a new approach.

Marijane says she's not even clear on what the PE group is trying to do. Damaris says one aspect is trying to get away from Java because a lot of new developers are not well versed in it. (Marijane's response, oh wow, because the semantic web toolchain is still really entrenched in Java.) Ralph notes that we don't really have a path to get developers into the VIVO code. Ralph agrees with the PE group's goal to bring in modern web development, but that there is still a role for Java here. Java can still be used as a backend technology to handle the semantics. There is a lot of stuff here that needs to be sorted out.

Marijane mentioned the larger context that VIVO exists in, the NIH/CTSA declaration that the ISF was the data standard that CTSA sites would be required to implement their data in, if VIVO moves away from that, does that hurt institutions' ability to adhere to that, and does anyone still care about that? Ralph says people do care at his institution. Brian reiterates what Damaris said about this being the core idea of VIVO and that the goal once was for other projects to build upon the model, and that there was a lot of excitement about Harvard Profiles using it, and not just having one software application that uses it. But the ontology should be what ties all these things together.

Ralph says he'd like to see a good and compelling ontology not just because it describes data, but because it describes data well. The VIVO project should be an example of how the ontology works. It's possible that project leadership has lost sight of this.

Marijane points out that the ontology does not yet describe data well, and Ralph agrees, for example it's missing description of Humanities scholarship. This group's goal should be to identify those gaps and to work towards making the ontology the best it can be so that the VIVO software can be the best example of using it.

Marijane raises the idea of wrapping ontologies the way you might wrap third-party libraries in software development, to ultimately insulate people from changes in the data model, inspired by a conference talk she recently watched (see here). Ralph agrees that we should have something like that, and that the Product Evolution team might work toward such a goal. It would be great if we could do that with the pre 1.7 ontology.

Marijane is going to review the PE meeting notes, and look at attending their meetings. Ralph suggests we work with Andrew to coordinate getting the various groups working together.