Concept mining for distributed alliance in multi-agent based virtual enterprises

Complex system manufacturing, such as that which takes place in the aerospace and automotive industries, is becoming increasingly dependent on distributed product development and is increasingly being conducted within the context of globally hosted virtual enterprises. These involve dynamic alliances and are characterized by a particular lifecycle: from partnership seeking, partnership cooperation to partnership dissolution. One of the basic problems in a distributed alliance is how to get an agent to find the best competent individual as a partner for a given task. The semantic Web and its associated new standards appear very promising as candidates to support a new generation of virtual enterprises based on heterogeneous systems integration and distributed alliances. Whilst knowledge can be represented in a machine interpretable way using the semantic Web, an agent can support automatic partner searching and documents processing in the Web. This paper proposes a matchmaking scheme for allocating tasks among agents based on the announcements of tasks and the capability of agents in DAML, where agents advertise their skills and local knowledge in a community and a task initiator seeks out the best-matched partners to form a dynamic alliance. Due to the autonomy of agents, the announcements are often from the agent's local perspective and the semantics of the announcements must be taken into account. To address this issue, a concept-mining scheme is then presented to reveal the actual meaning behind the announcements based on the ontology definition of a community. Consequently, distributed alliances in a virtual enterprise can take place in an automated/semi-automated way