Techniques for Partial Specification and Specification of Switching Systems
VDM '91 Proceedings of the 4th International Symposium of VDM Europe on Formal Software Development-Volume I: Conference Contributions - Volume I
Semantic acquisition in TELI: a transportable, user-customized natural language processor
ACL '86 Proceedings of the 24th annual meeting on Association for Computational Linguistics
Inferring formal software specifications from episodic descriptions
AAAI'87 Proceedings of the sixth National conference on Artificial intelligence - Volume 1
Hi-index | 0.00 |
Machine-assisted language translation systems for technical documents, guide humans through a process of selecting and composing variant partial translations. The constrained nature of technical sublanguages makes language processing aids cost-effective to build and use. Analogously, we have developed KITSS, a knowledge-based translation system for converting informal English scenarios of the desired behavior of complex reactive systems into formal, executable test scripts. A trainable parser and reference resolver capture domain-specific linguistic knowledge. A logic analyzer establishes coherence in the translation process in a role comparable to a "story understander". It checks the consistency of each step of a translated test script using a theorem prover, a planner, and logic-encoded background knowledge about the system under test. This helps correct common but serious specification errors, including underspecificity, omitted steps, and even some outright mis-statements. To evaluate how well such technology can scale, we have exercised our technology progressively on a graduated corpus of 100 behavior scenarios spanning 7 advanced calling features for a private telephone switch (PBX), successfully translating 70% into test scripts without any manual post-hoc editing. Our experience with KITSS has enabled us to identify many of the tradeoffs in accommodating informality in specification, versus demanding formality from a human agent.