Output details
11 - Computer Science and Informatics
University of Edinburgh
Experiences with Planning for Natural Language Generation
<22> Originality: The first study to empirically analyse the performance of modern automated planners on two problems in natural language generation. This work identified situations where the standard planning preprocessing techniques used in modern planners failed, and proposed two benchmarks for testing new planning approaches.
Significance: This is a journal-length version of a paper that was awarded Best Paper at SPARK 2008, the top international workshop on applications of automated planning.
Rigour: Descriptions of the testing domains used to generate the paper's empirical results were made available to the planning community and proposed as challenge problems.