A plan-based intelligent assistant that supports the software development
SDE 3 Proceedings of the third ACM SIGSOFT/SIGPLAN software engineering symposium on Practical software development environments
Probing limits to automation: towards deeper process models
ISPW '88 Proceedings of the 4th international software process workshop on Representing and enacting the software process
| Hi-index | 0.01 |
IN OPEN WORLD APPLICATIONS OF PLAN RECOGNITION, THERE IS USEFUL INFORMA- TION ABOUT THE STATE OF THE WORLD THAT CANNOT BE DIRECTLY OBSERVED FROM THE ACTIONS PERFORMED. WITHOUT THIS INFORMATION, SUBTLE ERRORS GO UNDETECTED, PREDICTIONS OF FUTURE ACTIONS LACK PRECISION, AND COMPETING INTERPRETATIONS CANNOT BE DISAMBIGUATED. ONE SOLUTION TO ACQUIRING ADDITIONAL STATE INFOR- MATION IS TO USE DOMAIN KNOWLEDGE TO MAKE PLAUSIBLE ASSUMPTIONS ABOUT THE MISSING VALUES USING THE OBSERVABLE VALUES. WE PRESENT A PLAN RECOGNITION ARCHITECTURE (BASED ON THE HIERARCHICAL PLANNING PARADIGM) THAT INCORPOR- ATES THIS NEW TYPE OF DOMAIN KNOWLEDGE: KNOWLEDGE THAT IS IMPRECISE, SUPPORTING CONCLUSIONS THAT ARE CONJECTURAL. `CREDIBILITY'', THE DEGREE OF AGREEMENT BETWEEN AN INTERPRETATION AND CURRENT `ASSUMPTIONS'' ABOUT THE WORLD, PROVIDES THE BASIS FOR IMPROVED ERROR DETECTION, PREDICTION, AND DISAMBIGUATION. THE ADDITIONAL DISCRIMINATION POWER IS FLEXIBLE. IN EXCEP TIONAL SITUATIONS AN `EXPLANATION'' FOR AN INTERPRETATION CAN BE GENERATED, CITING ASSUMPTIONS THAT MUST BE REVISED TO GIVE IT CREDIBILITY. THIS EXTEN SION TO PLAN RECOGNITION ENABLES DEEPER MODELING OF OPEN WORLDS.