| Abstract
|
Techniques that have traditionally been us … Techniques that have traditionally been useful for retrievingsame-domain analogies from small single-use knowledge bases, such asspreading activation and indexing on selected features, are inadequatefor retrieving cross-domain analogies from large multi-use knowledgebases. Blind or near-blind search techniques like spreadingactivation will be overwhelmed by combinatorial explosion as thesearch goes deeper into the KB. And indexing a large multi-use KB onsalient features is impractical, largely because a feature that may beuseful for retrieval in one task may be useless for another task. Thisthesis describes Knowledge-Directed Spreading Activation (KDSA), amethod for retrieving analogies in a large semantic network. KDSAuses task-specific knowledge to guide a spreading activation search toa case or concept in memory that meets a desired similarity condition.The thesis also describes a specific instantiation of this method forthe task of innovative design.KDSA has been validated in two ways. First, a theoretical model ofknowledge base search demonstrates that KDSA is tractable forretrieving semantically distant analogies under a wide range ofknowledge base configurations. Second, an implemented system that usesKDSA to find analogies for innovative design shows that the method isable to retrieve semantically distant analogies for a real task.Experiments with that system show trends as the knowledge base sizegrows that suggest the theoretical model's prediction of largeknowledge base tractability is accurate. geknowledge base tractability is accurate.
|
