Welcome to the home page for The Virtual Secretary
project. The project is partly support by the Research Council of Norway, Distributed
IT Programme (DITS), grant no. 112577/431 (through the Global Distributed Diary project).
Project summary
The Virtual Secretary project includes two phases: the first phase (ViSe)
focuses on user model-based software agents for information filtering and agent control
propagation; the second phase (ViSe2) concentrates on information integration via
co-operative agents in a distributed environment.
ViSe1
The user model is the core part of an adaptive
user interface. A user model contains information of the user’s past and current
tasks, and enables personification of the user interface and application software. In
order to respond to new actions, an adaptive user interface needs to detect what kind of
action the user is executing, and how these respond to previous user actions. This calls
for a relatively detailed user model capable of responding to extended user information
needs. In the ViSe1 project, we have used this feature to construct user model based
software agents. Through the propagation of user models to hosts where the rest of the
agent software (i.e., the body process) are installed, we have constructed a software
agent that both can exploit the features of mobile software agents, and be controlled by
an adaptive user with known behaviour. The user can delegate tasks to virtual secretaries
that operate on behalf of the user. A virtual secretary is supposed to assist and imitate
the user's actions, e.g., by letting virtual secretaries locate and filter information.
For a virtual secretary to be able to assist and imitate the users' actions, it needs some
knowledge about the user or the way that the user performs his tasks. This knowledge about
the user is closely related to the concept of a user model known from adaptive user
interfaces. The Virtual Secretary abstraction is the result of combining the agent
paradigm and the user model concept.
To illustrate the use of user model based software agents, we
have implemented the Virtual Library Secretary. The Virtual Library Secretary is able to
perform information retrieval and filtering, and remote file search on behalf of a user.
Experiments show that such agents are able to identify and adapt to patterns in a user
evaluated information stream. The agents also act on behalf of the user in the sense that
they remove irrelevant documents from the information stream presented to the user. By
delegating the evaluation of information to virtual secretaries and moving the evaluation
to the location of information, it is possible to reduce the use of network bandwidth. The
amount of network bandwidth is given by the agent’s precision level, i.e., amount of
relevant information divided the total amount of data retrieved.
ViSe2
In the ViSe2 project focuses on the construction of a multi-agent coopertaion
system. As a research vehicle, we have chosen to build intelligent agents that perfor
secretarial tasks for their users either by themselves or via cooperation. An individual
ViSe2 agent has limited knowledge and problem-solving capabilities. Therefore, the agent
has to interact with other agents to solve complex problems. In this sense, the agent's
ability to reason about the other agents' activities and thus find the peer becomes a key
issues. In the ViSe2 project, we have development a twin-base agent modeling
approach for ViSe2 agents to model and reason the activities of each other, and
thus efficiently locate peer agents for cooperation.
History
The ViSe project was proposed and founded in 1995, and is now headed by
Professor Gunnar Hartvigsen. The original
focus of the project was to construct a secure environment for software agents for
secretarial tasks in a global network [Hartvigsen
1995]. In early 1995, the project mainly addressed security and user-models [Hartvigsen, Helme, & Johansen 1995]. Later,
the project was extended to study cooperation technology [Hartvigsen et al. 1996] in a
distributed environment.
In 1996, the ViSe project was officially divided into two sub-projects (ViSe1
and ViSe2). The first phase (always called ViSe1) focused on information
filtrating and process migration. The second phase (ViSe2), addressed intelligent
software agents and efficient cooperation in Distributed Artificial Intelligence (DAI)
area and Multi-Agent Systems (MAS).
In the last periode of the project (1998-2000), the project mainly focused on the integration of software agents into a
Global Distributed Diary.
Parts of the ViSe project have been continued in the
DiPato and PaSent projects at the Norwegian Centre for telemedicine,
University Hospital of North Norway & Department of Computer Science,
University of Tromsų.
