OwlExporter 3.1 ---not getting temporary annotations

Hello,
I've run the ANNIE OrthoMatcher and OwlExporter Demo against a small corpus (I just want to obtain a list of the entity names and geolocations found in the corpus). But the temporary annotations are not appearing alongside the ANNIE annotations. Would appreciate any advice on this issue.

Thanks!
-Tony

More precisely, I'm not sure

Submitted by tonydaggett on Thu, 2012-09-20 09:55.

More precisely, I'm not sure how to construct the ontologies the program needs to generate the temporary annotations (ie, OwlExportClassDomain, etc). It doesn't say specifically how to accomplish this in the documentation.

Am I manually creating the ontologies in the gate\application-resources\ontologies folder? The documentation, otherwise very good, isn't clear on this.

Thanks,
-Tony

[5]

Re: OwlExporter 3.1 ---not getting temporary annotations

Submitted by rene on Mon, 2012-09-24 16:25.

Hi Tony,

The workflow is like this: you need an existing ontology that you want to populate. For the example pipeline, these are the domain.owl and nlp.owl (the first you need to replace for different domains). The output from OwlExporter is then stored in domain_out.owl and nlp_out.owl, so these are automatically generated.

If your goal is to construct the domain ontology itself (semi-automatically), you need an ontology learning system, not an ontology population one like OwlExporter.

For the population, you need to create the annotations that map annotations detected by your pipeline to ontology classes, properties, etc. These are the OwlExportClass etc. annotation you see in the demo pipeline. You have to create these yourself, e.g., using JAPE rules. These annotations need to be in place before the OwlExporter PR is run, since it is reading them as input for the population step.

So, if you want to integrate OwlExporter into your own pipeline, you need:

a domain ontology (you can generally simply reuse the NLP ontology, nlp.owl)
a PR (e.g., JAPE rules) to create the mappings (OwlExportClass etc.) between your annotations and the domain/nlp ontologies (see the example rules included with the demo pipeline and the documentation for their format)

Hope that helps,

René