ITRI-98-16
Kees van Deemter

"Retrieving Pictures for Document Generation"
Also published in Proc 14th Twente Workshop on Language Technology, pp 117--128

This paper sketches how picture retrieval can be used by a document
generation system for the inclusion of `photographic' pictures in the
documents generated by the system. The exposition is based on the What
You See Is What You Meant ('wysiwym') approach to knowledge
editing and document generation (Power and Scott 1998, Scott et
al. 1998). The paper describes an algorithm that makes use of a
library of pictures, each of which is associated with a set of logical
representations to facilitate retrieval. The paper introduces the twin
notions of pictorial underspecificity and pictorial overspecificity,
and focuses on the problems stemming from these two phenomena.