Agents can effectively host an architecture to support a digital library and be present in the electronic publishing process (process based on a evolution of the required phases mentioned for the traditional publishing of a scientific paper as described in ), helping to bridge the gap between information producers (i.e. authors), publishers and information consumers. This poster presents a Virtual Research Digital Library (VRDL) based on a architecture of agents and describes the on-going validation process.
The architecture is composed of eight different types of agents (see Figure 1).
Here is a description of the types of agents and their roles (more details in ):
The proposed architecture has been developed with the ZEUS
agent system from British Telecom, which is open source
The VRDL has two validation cycles:
The first validation cycle involves on-going debugging processes and a technical verification procedure. This technical verification procedure is being accomplished through verification questionnaires to the VRDL administrator and tracking (i.e. statistical analysis produced by ZEUS and log file analysis).
The second validation cycle requires that a sufficiently large sample of users in real-life situation provide information on user friendliness and similar issues, as well as testing the feasibility of the system when used on large scale. The verification procedure of this second validation cycle will be accomplished through verification questionnaires to the VRDL administrator and different simulated types of users, and tracking.
The first cycle has ended (except for the debugging part, which continues through the validation process) with good results in terms of overall system functionality. Some issues regarding specific agent's performance and minor bugs are being corrected. The second validation cycle in currently under way.
It is clear that VRDL shows how agents can efficiently be used to host a digital library and the advantages that result of its application. The proposed system has been implemented through a prototype and after the first validation cycle one can conclude that it performs well. Future concerns are the increased efficiency possible, and giving feedback to user requirements within the proposed architecture.