An Automatic Instance Expansion Framework for Mapping Instances to Linked Data Resources

Abstract

Linked Data is an utterly valuable component for semantic technologies because it can be used for accessing and distributing knowledge from one data source to other data sources via structured links. Therefore, mapping instances to Linked Data resources plays a key role for consuming knowledge in Linked Data resources so that we can understand instances more precisely. Since an instance, which can be aligned to Linked Data resources, is enriched its information by other instances, the instance then is full of information, which perfectly describes itself. Nevertheless, mapping instances to Linked Data resources is still challenged due to the heterogeneity problem and the multiple data source problem as well. Most techniques focus on mapping instances between two specific data sources and deal with the heterogeneity problem. Mapping instances particularly relying on two specific data sources is not enough because it will miss an opportunity to map instances to other sources. We therefore present the Instance Expansion Framework, which automatically discover and map instances more than two specific data sources in Linked Data resources. The framework consists of three components: Candidate Selector, Instance Matching and Candidate Expander. Experiments show that the Candidate Expander component is significantly important for mapping instances to Linked Data resources.