The Semantic Web and Human Inference: A Lesson from Cognitive Science

Abstract

Abstract. For the development of Semantic Web technology, researchers and developers in the Semantic Web community need to focus on the areas in which human reasoning is particularly difficult. Two studies in this paper demonstrate that people are predisposed to use class-inclusion labels for inductive judgments. This tendency appears to stem from a general characteristic of human reasoning -using heuristics to solve problems. The inference engines and interface designs that incorporate human reasoning need to integrate this general characteristic underlying human induction.

Searching Linked Objects with Falcons

“…The Semantic Web brings more structured data with rich semantics, which cannot be satisfactorily utilized by a purely keyword-based search engine to serve object search. Cognitive science has shown that people are predisposed to use typing information rather than other property information to perform human reasoning (Yamauchi, 2007), whereas typing information (rdf:type) is also widely used by data producers. Therefore, it is practical and also feasible to exploit typing information in the system to improve object search, for example, enabling to filter the resulting objects by specifying the type of objects being sought for.…”

Integrating Lightweight Reasoning into Class-Based Query Refinement for Object Search

“…Compared to the hypertext Web, the Semantic Web brings more structured data with rich semantics, which cannot be satisfactorily utilized by a purely IR-based search engine to serve object search. Recently, cognitive science shows that people are predisposed to use typing information rather than other property information to perform human reasoning [16], while typing information ( rdf:type) is also widely used by data producers. Therefore, we exploit typing information in the system to improve object search.…”