Copy the page URI to the clipboard
Nguyen, Tu Anh T.; Power, Richard; Piwek, Paul and Williams, Sandra
(2012).
URL: http://nlp.cs.uic.edu/inlg2012/
Abstract
A useful enhancement of an NLG system for verbalising ontologies would be a module capable of explaining undesired entailments of the axioms encoded by the developer. This task raises interesting issues of content planning. One approach, useful as a baseline, is simply to list the subset of axioms relevant to inferring the entailment; however, in many cases it will still not be obvious, even to OWL experts, why the entailment follows. We suggest an approach in which further statements are added in order to construct a proof tree, with every step based on a relatively simple deduction rule of known difficulty; we also describe an empirical study through which the difficulty of these simple deduction patterns has been measured.