Séminaire du 15 Février 2008


LIPN: comment y aller


10h00-11h00: Jörg Desel, KU Eischtätt, Allemagne, "From Human Knowledge to Process Models"

Abstract: in this talk I suggest a novel approach for a systematic generation of a process model in an informal environment. The approach is based on the claim that the knowledge about the process to be
modelled is distributed in several involved people's minds. Some people have knowledge about the general process where the single activities are on a high level of abstraction and have to be refined. Other people only know something about some detail of the process, i.e., about the refinement of an activity of the general process which defines a subprocess. Moreover, it is assumed that these domain experts can more easily define instances, i.e. runs, of the general process (of a subprocess, respectively) than the process itself. The approach employs new techniques developed for process mining and Petri net synthesis and adapts these techniques to generate processes from example runs. It is based on a formal definition of partially ordered processes, which allows to proceed in a modular way: Subprocesses and general processes are generated independently. Finally, it is argued that the approach is a suitable first step in a general method for process definitions which is followed by validation techniques.

11h00-11h30 : Olivier Bertrand, PhD student at LIPN and ONERA, "Towards a coloured Petri nets semantics for a chronicle language"

Abstract: chronicles describe relationships between events in a sequence ordered in time, and are used to model activities that should be detected within an observed events flow. A chronicle language provides a syntax for the different chronicle operators considered. In this work, coloured Petri nets are used to model the recognition of a chronicle within a flow of events, and the occurrence of an event to be detected is modelled by the firing of the corresponding transition. We provide coloured Petri nets to model the (recognition of ) chronicles expressed with logical and temporal operators, as well as minimum and maximum time delays. We show how the composition of operators can be modelled by a composition of the coloured subnets associated with the different operators. This composition may be achieved either through transition substitutions, or through place fusions. In both cases, the algebraic properties of the operators are reflected in the coloured nets. The execution of the resulting coloured net on an event flow show that the various occurrences of a chronicle are indeed recognised. This work thus provides elements to model complex chronicles recognition, and to define a semantics of the chronicle language in terms of the used subset of coloured nets.