Address extraction: a graph matching and ontology-based approach to conceptual information retrieval

Address and related location-awareness information can be retrieved and extracted from the Web using content-based IR technologies. Much recent research on content-based information retrieval focuses on conceptual analysis of unstructured texts on the Web. This paper illustrates an address extraction application to achieve an ontology-based conceptual IR system with graph matching. Our key idea is that a document can be represented as a sub-graph of a predefined ontology graph. An approximate graph matching approach is used for content (address) extraction. This work is part of an ongoing project to develop an intelligent search agent to support driving-related information extraction from the Web.