search

laurentromary / stdfSpec

Specification of a stand-off element for the TEI guidelines
12 stars 7 forks source link

need <ref> or <annotationRef>; <linkGrp> would make sense as well #17

Open bansp opened 7 years ago

bansp commented 7 years ago
  1. I am taking a very analytic approach to creating and assembling annotations, and I found myself in need of expressing reference in a clean way (not by abusing <seg>, which allows for <ref> children), right under listAnnotation.

What I have is something like:

<listAnnotation type="segmentation" xml:id="lA1">...
<listAnnotation type="terminals" xml:id="lA2">...
<listAnnotation type="nonterminals" xml:id="lA3">...

-- these are in a way general-purpose annotations, especially the upper two (and note that I can have many instances of each, if I want to support multiple annotations for a single resource), and now I would like to follow up with something like:

<listAnnotation type="tree">
   <annotationRef type="root" target="lA3.5"/>
    <annotationRef type="nonterminals" target="lA3"/>
    <annotationRef type="terminals" target="lA2"/>
 </listAnnotation>

... to recreate the typical Tiger-XML-like serialization. I could do with <tei:ref> here, although <annotationRef> appears to me to carry the right amount and the right kind of semantics, as opposed to the extremely general <ref>.

  1. Additionally, it's struck me now, belatedly, that alignment was after all among the original goals behind the standoff approach (recall: cesAna and... cesAlign), so it would not be in any way shameful, and indeed quite sensible, to re-implement the second part of the original goal in an approach that is, among others, meant to cater to people who would otherwise attempt to revive the CES in some form. In other words, I would like to suggest adding <tei:linkGrp> as a child of listAnnotation.

These are related proposals, but with definitely different scopes. <linkGrp> points outside (e.g. for sentence alignment in a multilingual resource, etc.), whereas the putative <annotationRef> would be a <standOff>-internal pointer.

bansp commented 7 years ago

Ah, to be sure: I'm not asking for work to be done, I'm just asking for discussion or green light :-)

bansp commented 7 years ago

Ping. I'm moving in this direction... am I on my own or is there interest in enriching the standOff model with this? (hint: there should be interest in that ;-))