Collaborative Understanding of Distributed Ontologies in a Multiagent Framework: Experiments on Operational Issues

Abstract

This chapter describes a set of experiments that uses a multiagent framework for collaborative understanding of distributed ontologies (CUDO). Our current focus is on the operational issues of such collaboration among the agents, with each agent managing an information database in a distributed information retrieval simulation. To facilitate collaborative understanding, each agent maintains an ontology and a translation table with other neighboring agents to map between each own concepts and the neighbors’. Based on an infrastructure prototype, our experiments have focused on how neighborhood profiling, the translation tables, and query experience impact the collaborative activities among the agents. The specific objectives of our analyses are to investigate (a) the recognition of useful neighbors for sharing queries, (b) the efficiency of query handling in different real-time scenarios and with different resource constraints (such as the number of threads and available translations), and (c) the effects of different concepts and query demands on collaborative understanding. Our results show that the different resource constraints influence the collaborative activities significantly and thus also impact how the agents learn of each others ontologies.