Re: named entity annotation / certainty / provenance

Hi James,

On Thu, Aug 16, 2012 at 9:28 AM, James Chartrand <jc.chartrand@mac.com>wrote:

> We're annotating named entity references (to people, places, events,
> organizations) in text like so:
>
> <AnnoA> a oa:Annotation ;
>    oa:hasTarget <urn:uuid:CB37E17E-B1EF-48CE-B098-0CE8FAA86A6D> ;
>    oax:hasSemanticTag <http://dbpedia.org/page/MrJones> ;
>    oa:annotator <http://somewhere.ca/people/JamesChartrand> ;
>    oa:annotated "2012-08-15T12:10:54Z" ;
>    oa:generator <someSoftwareURI> ;
>    oa:generated "2012-08-15T12:10:54Z" ;
>    oa:modelVersion <http://openannotation.org/spec/core/20120509> .
>
>   <urn:uuid:CB37E17E-B1EF-48CE-B098-0CE8FAA86A6D> a oa:SpecificResource ;
>     oa:hasSource <http://somewhere.ca/work/345345> ;
>     oa:hasSelector <urn:uuid:2ADFF7EE-AB19-4BA3-94EC-55EE0BA645C2> .
>
>   <urn:uuid:2ADFF7EE-AB19-4BA3-94EC-55EE0BA645C2> a oax:TextOffsetSelector
> ;
>     oax:offset 244 ;
>     oax:range 7 .
>
>   <http://somewhere.ca/work/345345> a dctypes:Text .
>
>  Three questions:
>
> 1.  Does this seem right?
>
>
At first glance I would say yes.


> 2.  We'd like to assert the confidence with which we're annotating -- how
> certain we are that this reference is to the given person.  Are there any
> recommendations for asserting certainty?
>

I would suggest to create your own property or mechanism and then
contribute it back to the group. That is certainly something others are
doing. In my project I have a score property (score that is coming from
text mining algorithms) attached to the Annotation node. It would be
interesting to collect the different use cases on this topic on the wiki.


>
> 3.  Should the oa provenance predicates be used in preference to the PROV
> ontology (http://www.w3.org/TR/prov-o/) or can PROV be used
> interchangeably?  Is PROV recommended for provenance predicates not covered
> by OA?  Any plans to move to PROV entirely?
>

We worked out some mappings
oa:annotator subPropertyOf prov:wasAttributedTo
oa:annotated subPropertyOf prov:generateAtTime
oa:generator subPropertyOf prov:wasAttributedTo
oa:generated subPropertyOf prov:generatedAtTime.

I believe we will include/publish them at one point. However, as you can
see, the expressiveness of the prov properties is not the same.

Best,
Paolo