Meeting Notes 20101018 NEMO core meeting with Paea & Mariano - Troubleshooting RDF data classification

From Nemo

Jump to: navigation, search
  • Convened by: NEMO core
  • Presenter: Paea LePendu
  • Notes by: Gwen Frishkoff
  • Where: WebEx
  • Attendees: Gwen, Paea, Mariano, Dejing, Bob, Snezana

Contents

Agenda

The goal for this meeting was for Paea to present some ideas about incremental testing of ontology-based classification of RDF data in Protege.

Minutes

Paea demontrated that HermIT works o.k. on ontology itself in Protege

But then Paea opened RDF and imported NEMO.owl and showed that classification did not work ("inconsistent ontology" error)

What now? Troubleshoot using Explainer (special version of Protege that we looked at last time)

Bob pointed out that he has tried out two different ways to represent typed literals. Wondering whether one representation may work better than the other. Mariano checked and found that the RDF representation is correct as currently specified.

Gwen suggested to define individuals in Protege, check classification results, then export to RDF and examine code.

Note that rules specify datatype as integer.

Updating RDF

Mariano suggested to restructure the RDF to make it easy to debug. Or maybe just open Bob's RDF output in Jena and save in more human-readable format.

Debugging workflow -- make up small set of test cases, articulate incremental set of steps.

Paea asked about size of dataset. Even grand average data representation consists of 18k RDF triples.

SPARQL is NOT RDB or ontology database Either of these may be an option, as well (if we need them).

Gwen mentioned concern about whether equiv class syntax is expressive enough to encode complete pattern rule. If not, could we use SWRL? Maybe, but in that case, we need a rule-based reasoner on back-end. Note that Pellet works with SWRL (Mariano) Try Pellet API and use it to load data & rules. (Not sure if it will work within Protege)

We agreed to first try to specify complete rule in OWL. If we conclude that OWL syntax is not rich enough, the next step may be to research SWRL.


Action items

  • Gwen talk with Paea -- give complete representation of rule and talk about process of coding it in OWL
  • Gwen, Bob assemble test dataset that is stripped down version of simERP RDF data (with known classification results for debugging)
  • If OWL syntax is not sufficiently expressive, discuss how to proceed with SWRL and look into rule-based reasoners that can work with SWRL -- inside or outside Protege.

WebEx Recorded Archive

If you wish to review the recording of the web conference, click the link below:

[[1]]

NEMO Teleconference-20101019 1717-1 Tuesday, October 19, 2010 2:15 PM New York Time 56 minutes

/wiki/images/1/17/Fish1.png /wiki/images/e/ea/Fish2.png /wiki/images/f/fa/Fish3.png /wiki/images/f/ff/Fish4.png /wiki/images/4/40/Fish5.png /wiki/images/c/c5/Fish6.png
Personal tools