Knowledge-intensive processes: new ideas or "old wine in new bottles”? Role of Knowledge Representation Capture at the conceptual level the domain of interest • Abstract away form implementation/operational issues • Build a formal model of the domain of interest that allows for formal reasoning • Enabling, if possible, automated reasoning • Giuseppe De Giacomo
Knowledge-intensive processes: new ideas or "old wine in new bottles”? Role of Knowledge Representation: DATA Capture at the conceptual level the domain of interest • UML class diagrams/ER diagrams • Abstract away form implementation/operational issues • Abstract from operational data used in DBMS/Programs • Build a formal model of the domain of interest that allows for formal reasoning • Nowadays Description logics capture UML class diagrams/ER diagrams! • Enabling, if possible, automated reasoning • DL-lite! • Today, after 30+ years, we finally have tools that are mathematically well founded and really practical!!! Giuseppe De Giacomo
Knowledge-intensive processes: new ideas or "old wine in new bottles”? Role of Knowledge Representation: PROCESSES without DATA (or “propositional”) Capture at the conceptual level the domain of interest • BPMN, UML Activity Diagrams, Petri Nets, Transition Systems (finite) • Abstract away form implementation/operational issues • Abstract version of the program/process. Maybe event too much. Data almost • fully abstracted away! Build a formal model of the domain of interest that allows for formal reasoning • Petri Nets, Transition Systems are mathematical structures! • Enabling, if possible, automated reasoning • Model Checking! • Giuseppe De Giacomo
Knowledge-intensive processes: new ideas or "old wine in new bottles”? Role of Knowledge Representation: PROCESSES + DATA Capture at the conceptual level the domain of interest • BPMN, UML Activity Diagrams, (Infinitely) Colored Petri Nets, • (Infinite)Transition Systems - Data are not represented explicitly AI/KR: Dealing with data and process nontrivial: Frame Problem, Ramification • Problem, Qualification Problem – almost ignored in SE/BPM literature See [Borgida, Mylopoulos, Reiter: On the Frame Problem in Procedure • Specifications. IEEE Trans. Software Eng. (1995)] AI/KR: Specify tasks a la PDDL? • Abstract away form implementation/operational issues • Yes • Build a formal model of the domain of interest that allows for formal reasoning • Yes • Enabling, if possible, automated reasoning • Needs more work! But, see EU ACSI results on verification! • Giuseppe De Giacomo
Recommend
More recommend