Re: [agtk-devel] Spec for discontinuous annotation

[Steven B. <sb...@cs...>]

Hi Nikhil,

On Mon, 2004-05-17 at 06:01, Nikhil Dinesh wrote:
> More abstractly the arguments are multiple partially-overlapping
> discontinuous spans of text and I was wondering:
> 
> a. What is the ag recommendation for such spans? There seem to be
> may ways to go about it.

The most extended discussion of the use of AGs for annotating discourse
and dialogue is in:

@inproceedings{BirdLiberman99dtag,
  author={Steven Bird and Mark Liberman},
  title={Annotation graphs as a framework for multidimensional
linguistic data analysis},
  year=1999,
  booktitle={Towards Standards and Tools for Discourse Tagging --
Proceedings of the Workshop},
  publisher={Somerset, NJ: Association for Computational Linguistics},
  pages={1--10},
  note={[xxx.lanl.gov/abs/cs.CL/9907003]}
}
                                                                                In general, annotate higher-order relations between text spans by employing labels which reference other annotations.

> b. How does ag represent syntactic traces? As self-loops?

No, since AGs are acyclic.  Instead, you set up two anchors having the
same offset.  The trace annotation connects these anchors.

> c. If I do convert it to the spec, is there a display and query part
> of the toolkit that would work on it?

In short, no.  Each of the existing tools (MultiTrans, TableTrans,
TreeTrans, InterTrans, ...) makes assumptions about the data which are
likely to be incompatible with your model.  However, there's a chance
that more recent applications developed at LDC, not currently
distributed with the toolkit, would meet your needs.

-Steven Bird