Re: [uima] [Type System Base Model subgroup]

[Karin Verspoor <verspoor@lanl.gov>]

Title: Re: [uima] [Type System Base Model subgroup]

Thanks to all for the feedback so far.

I’ve pulled my initial comments, adjusted on the basis of subsequent discussion, as well as the various suggestions from other respondents, into the attached document.  Perhaps we can continue with our discussion based on the points raised there.


Some more thoughts on the Annotations tied to spans issue:

- an advantage of having a subtype of Object that encompasses both “span”-grounded markup and markup not tied to spans is that any properties that might apply to both can be shared.  I’m thinking specifically of Provenance information here (which I added to the list of open issues), or the weight/relevance score that Pascal suggested.  In an implementation I recently worked on, we had to distinguish between “ProvenancedAnnotations” and “ProvenancedFeatureStructures” because there was no way to add the feature “provenance” to a common supertype.

- A problem with having RegionalReferences that span an entire document, in particular as the preferred way to represent annotations/metadata that is not tied to a specific span, is that this makes such Annotations difficult to handle in a visualization – what does it really mean to highlight the entire document as being relevant to e.g. a keyterm?  I think it would be preferable to have an explicit distinction between markup tied to a span and markup not tied to a span.

- What about having the following hierarchy:

Object
  |
Metadata (sofa: SofaReference)
 |
Annotation (regionalRef: RegionalReference)  <or a list of references, see OI.4.

So, Metadata such as keyterms or Entities not tied to specific document references are still associated with a sofa, and any other features (e.g. provenance, relevance) generic to any markup can be added there.

Karin

_______________________________________________________________
Karin Verspoor, Computational Linguist
Knowledge and Information Systems Science team
Computer, Computation & Statistics division
http://public.lanl.gov/verspoor
email: verspoor@lanl.gov   Mail: Los Alamos National Laboratory
phone: 505-667-5086              PO Box 1663, MS B256
fax:   505-667-1126              Los Alamos, NM 87545
_______________________________________________________________

TypeSystemBaseModel.doc