Skip to main content

An Architecture for Interactive Problem Solving by Cooperating Data and Knowledge Bases

Objective

The objective of ESTEAM was to design and implement an expert system architecture for advice-giving systems.
The function of an automatic adviser is to assist an inquirer with a problem. The problem may be ill-defined and the number of potential solutions may be large. In this situation, the adviser can help the inquirer to provide a statement of his or her goals and a description of the problem sufficient for the generation, by the machine, of trial solutions for consideration.
The objective of the project was to design and implement an expert system architecture for advice-giving systems. The function of an automatic adviser is to assist and inquirer with a problem. The problem may be ill defined and the number of potential solutions may be large. In this situation, the adviser can help the inquirer to provide a statement of his or her goals and a description of the problem sufficient for the generation, by the machine, of trial solutions for consideration. The architecture was divided into 2 complementary strands: for designing and implementing architectures for heterogeneous distributed advice-giving systems; and methods and tools to model knowledge in advice-giving expert systems. These were a dialogue manager, a problem solver, a cooperative answering agent and a database agent. The main computational problem tackled by the combined cooperative action of the agents was how to manage those complexities of advice-giving that require the integration of knowledge and data from a variety of sources. This was solved by controlling the cooperative functioning of several sources of knowledge by using different representational schemes interpreted by different inference engines. Each knowledge source was considered to be an independent agent, only communicating with other such agents via messages encoding queries and answers. In addition to architectural issues, the dialogue system was provided with the capability for modelling lines of thought of the user, as perceived through the person-machine interaction. A simplified financial investment adviser was constructed to provide a limited example suitable for an application study. Knowledge acquisition was completed, and the capability of pairs of actors to cooperate was tested by processing problems in this domain. A first demonstration showed the integration of the problem solver, the dialogue manager and the database agents. A second demonstration illustrated the integration of the cooperativ e answering agent, the dialogue manager, the rule base management and the database.
The architecture was divided into two complementary strands:
-The AGES architecture, dealing with concepts and tools for designing and implementing architectures for heterogeneous distributed advice-giving systems. This architecture has been ported from its development environment (TI Explorer) to a Sun/Unix envir onment.
-Methods and tools to model knowledge in advice-giving expert systems. These "cooperative agents", which each take control in turn, were a dialogue manager, a problem solver, a cooperative answering agent and a database agent (to ORACLE).
The main computational problem tackled by the combined cooperative action of the agents was how to manage those complexities of advice-giving that require the integration of knowledge and data from a variety of sources. This was solved by controlling the cooperative functioning of several sources of knowledge by using different representational schemes interpreted by different inference engines. Each knowledge source was considered to be an independent agent, only communicating with other such agents via messages encoding queries and answers.
In addition to architectural issues, the dialogue system was provided with the capability for modelling lines of thought of the user, as perceived through the person-machine interaction.
A simplified financial investment adviser was constructed to provide a limited example suitable for an application study. Knowledge acquisition was completed, and the capability of pairs of actors to cooperate was tested by processing problems in this domain.
A first demonstration showed the integration of the problem solver, the dialogue manager and the database agents. A second demonstration illustrated the integration of the cooperative answering agent, the dialogue manager, the rule-base management and th e database.
Exploitation
The major contribution of ESTEAM was the provision of a prototype architecture for heterogeneous distributed advice-giving systems. The results will be exploited internally by the partners for the development of specific applications.

Coordinator

CAP GEMINI INNOVATION
Address
118 Rue De Tocqueville
75017 Paris
France

Participants (4)

Centro Studi e Laboratori Telecomunicazioni SpA
Italy
Address
Via G. Reiss Romoli, 274
10148 Torino
Office National d'Études et de Recherches Aérospatiales (ONERA)
France
Address
2 Avenue Édouard-belin
31055 Toulouse
PHILIPS SA
Belgium
Address
Av. Albert Einstein
1348 Louvain-la-neuve
Politecnico di Milano
Italy
Address
Piazza Leonardo Da Vinci 32
20133 Milano