A Spreading Activation Framework for Ontology-Enhanced Adaptive Information Access within Organisations

摘要

This research investigates a unique Indexing Structure and Navigational Interface which integrates (1) ontology-driven knowledge-base (2) statistically derived indexing parameters, and (3) experts' feedback into a single Spreading Activation Framework to harness knowledge from heterogeneous knowledge assets. Within an organisation, organisational ontologies capture precise knowledge about organisational entities: people, projects, activities, information sources and so on. We extract useful entities and their relationships from an ontology-driven knowledge base. We also process collections of documents (archives) accumulated in heterogeneous information-bases within an organisation and derive indexing parameters. This information is then mapped to a weighted graph (spreading activation network). The network contains three distinct sets of nodes representing documents, ontological entities and statistically derived entities. Document nodes are connected to both ontology-driven entities and statistically derived entities, and vice-versa with relevant weights. Retrieval is performed by spreading query-based activation into the network and selecting the most-activated nodes. Experts as well as users in the organisation either navigate the network using associative relations among nodes or with specific queries. Expert's feedback is captured and the network weights are continuously adapted. This framework essentially combines precise knowledge (ontology-driven), non-precise knowledge (statistically driven) and Expert's feedback (adaptation and refining) into a single framework for effective information retrieval and navigation.