A Proposed Model for Effective Verification of Natural Language Generatio Systems

摘要

In this paper we proposed an approach for effective verification of NLG systems, based on conventional evaluation of language processing systems combined with software testing and system verification concepts drawn from the intelligent systems community. We presented a first step in the implementation of the verification strategy in the form of a specific strategy for the verification of lexicalized tree adjoining grammars (LTAGs) as used in natural language generation systems. In future work we will determine the usefulness of this approach by testing it on a number of LTAGs that have been used in conjunction with a specific language generation system.