Regression testing for grammar-based systems

In complex grammars, even small changes may have an unforeseeable impact on overall system performance. As grammar based systems are increasingly deployed for industrial and other large-scale applications, it is imperative to have systematic regression testing in place. Systematic testing in grammar-based systems serves several purposes. It helps developers to track progress, and to recognize and correct shortcomings in linguistics rules sets. It is also an essential tool for assessing overall system status in terms of task and runtime performance. This paper describes best practices in two closely related regression testing frameworks used in grammar-based systems: MedSLT, a spoken language translation system based on the Regulus platform, and a search and question answering system based on PARCs XLE syntax-semantics parser.

Citation

Chatzichrisafis, N.; Crouch, D.; King, T. H.; Nairn, R.; Rayner, M.; Santaholma, M. Regression testing for grammar-based systems. Proceedings of GEAF (Grammar Engineering Across Frameworks) Workshop; 2007 July 13-15; Stanford, California.