The system architecture under consideration is shown in Figure 1. Our cooperative KBS consists of a set of sources with local knowledge base system {KBSi, for 1 ≤ i ≤ m}, where each KBSi is a pre-existing autonomous knowledge management system contains knowledge data source. Each system provides a set of services to share and access its knowledge locally and remotely. For example each system provides acquaintance services for creating acquaintances with other systems, query service for retrieving information over the network, and update service for updating local and remote knowledge. Once a system joins the network, it can establish acquaintances with other systems exchanging their schemas and use services exists in local system and its acquaintances. An acquaintance is a connection between two systems and is established by generating mapping tables (Kementsietsidis et al., 2003) on both systems. The acquaintances are transient since each system is fully autonomous and joins or leaves the network at its own will. Formally, we can define our collaborative KBS as follows.
System model
Definition 1:A collaborative KBS is a pair (N,M). Here, N = {(K,L}) is an undirected graph, where K = {KB1, …,KBn} is a set of knowledge base system or simply called sources, L = {(KBi, KBj) |KBi, KBj ∈ K} is a set of acquaintances. Each source KBi is associated with a knowledge database with schema KDBi[Wi], and each acquaintance (i,j) is associated with a set Mij ∈ M of mapping tables.
At the beginning, when a source joins a collaborative knowledge base system, it is only aware of its acquaintance. Compare to the discovery protocol (Kantere, Kiringa, & Mylopoulos, 2003), actual knowledge update is different from two perspectives. 1) Update algorithm continues its operation until it reaches a fixed point while discovery algorithm stops when a source is reached twice. 2) Actual queries and propagation is not performed by the discovery algorithm.