The Internet offers remote access to many information systems to users independent of time and location. This paper describes an agent based approach to deal with issues that rise from the differences between user interfaces to functionally similar information systems. The approach yields agents that interact with user interfaces to extract the model of the underlying information system. The model is represented by labeling input fields of the interface with their meaning. The agents are modeled after human dialogue behavior and human computer interaction. The interaction agents are capable of independently and robustly querying the information system without explicit instructions. Two types of agents are implemented and tested on European sites of rail travel planners. The agent design and test results will be reported.
L. Ton, Léon J. M. Rothkrantz