Patrick Brézillon: Keywords and Objectives
Back to
Back to
Keywords:
Context-based
Intelligent Assistant Systems: Artificial Intelligence,
Decision Making Support
Context Modeling: Contextual Graphs,
Shared context, Proceduralized Context
Representation of the Knowledge,
Reasoning and expertise based on context
Contextual Management of Knowledge:
Acquisition, Learning, Configuration, Retrieval,
Contextual Management of
Communication: Cooperation, Explanation, Human-Machine
Interaction
Context-Sensitive Applications,
Dynamic Modeling of the User
Objectives
The Context-based Intelligent Assistant Systems (CIASs)
represent the generation after the Knowledge-Based Systems (KBSs)
and
decision support systems. The goal of a CIAS is to support
intelligently
users, not to replace them. In the applications on which I work
with my
students, the user is an operator. Operator and CIAS jointly solve
the
problem, but operators alone make the final decision about the
real-world process they have in charge. CIASs are developed in a
multi-agent spirit that permits a progressive development, each
agent
being the subject of a Ph.D. Thesis.
My research focus on four aspects of CIASs, namely cooperation,
explanation generation, incremental knowledge acquisition and the
notion
of context. These aspects are strongly correlated, as already
pointed
out for explanation generation and context (see Karsenty &
Brezillon, 1995). Now, I study:
- modeling of context through a real-world application for
the
subway (the SART project),
- context as the basis of any development of CIASs,
- context at the level of knowledge representation (the onion
metaphor),
- context at the level of the reasoning (contextual
knowledge,
procedralized context and their dynamics),
- consequences of integration of these aspects for CIAS
design
and development, and
- fusion of KBS and Decision Support Systems in the framework
of
the CIAS paradigm.
The SART project (since 1996):
- The goal is to develop a CIAS for supporting the operator
who
has in charge the regulation of a subway line when an incident
occurs.
Three agents are currently developed: a line configurator, a
traffic
simulator and an incident manager.
- This work is the object of a contract between the
University
and the Federal University of Rio de Janeiro (UFRJ, Brasil)
for the
realization of the SART project. In this framework, I was the
co-director of the thesis of Mr. C. Gentile and Mr. M. Secron,
Ph.D.
students at the FURJ and Engineers at Metrô, the subway
company
at
Rio of Janeiro. The subjects of their thesis are to develop
the traffic
simulator and the line configurator of a subway line.
- This work is also the object of a contract between the
University Paris 6 and RATP (company that has in charge the
subway in
Paris) for the study of an agent, namely the Incident Manager,
subject
of the Ph.D. thesis of Ilham Saker and Laurent Pasquier.
- Globally, around 40 persons are directly concerned (or have
been concerned) by the SART project.
- For more details, see (Brezillon et al., 1997) or here (in French).
The COSINUS project (1996-1999):
- The goal is the development of a CIAS for supporting the
Enologist in the predition of stop fermentation in vineyard
cellars.
- I co-directed the thesis of Mrs. Juliette Agabra for the
knowledge-based agent, the second part of the whole system
being a
neural network, subject of the thesis of Mr. Hugo Zaragoza.
- Eight persons were concerned by the COSINUS project.
- For more details, see (Agabra et al., 1997).