Emergent semantics in knowledge sifter: An evolutionary search agent based on Semantic Web services

This paper addresses the various facets of emergent semantics in content retrieval systems such as Knowledge Sifter, an architecture and system based on the use of specialized agents to coordinate the search for knowledge in heterogeneous sources, including the Web, semi-structured data, relational data and the Semantic Web. The goal is to provide just-in-time knowledge to users based on their decision-making needs. There are three important factors that can assist in focusing the search: 1) the user's profile, consisting user preferences, biases, and query history, 2) the user's context to focus on the current activity, and 3) the user's information space, in which he may receive the information on specialized hardware with limited bandwidth, implying mat the knowledge must be filtered and tailored to die presentation medium. Emergent semantics in me context of Knowledge Sifter allow for evolutionary adaptive behavior. We present a meta-model that captures the agent operation and interactions, as well as the artifacts that are created and consumed during system operation. These are stored in a repository, and a collection of emergence agents are presented that perform emergence functions such as: data mining for patterns; concept discovery and evolution; user preferences tracking; collaborative filtering of user profiles; results ranking; and data source reputation and trust.