Semantic Association Analysis in Ontology-Based Information Retrieval

Introduction

The current Web provides a universal platform to explore and contribute to the global information network. Undoubtedly, the Web has emerged as the world’s major information resource with immediate accessibility in a world-wide scale. Currently, in most of the cases, in order to transform available information into meaningful knowledge, machines have to depend on the human inference ability (Craven, DiPasquo, Freitag, McCallum, Mitchell, Nigam, et al., 2000). Contemporary popular online search engines and information retrieval systems index and search the Web documents based on analysis of the document link structures and keywords. The keywords are often extracted from the documents according to the frequency of occurrence and considered as standalone entities without application contexts and other semantic relationships. This superficial understanding of content prevents retrieving implicitly-related information in most of the cases. It also in some cases returns irrelevant results to the user. In the context of multimedia Web, the current search systems are even more limited. Most of the multimedia search on the current Web relies on text explanations extracted from accompanying pages or tags provided by content authors. There are commercially successful Web sites for multimedia publishing, sharing, and retrieval on the Web such as MySpace¹, YouTube², and Flickr³. These Web sites have demonstrated a great achievement in acquiring millions of users to form communities and contribute to content generation; they also provide interfaces to search and view the published contents, but the search functions regularly rely on conventional methods and keyword matching mechanisms. As overwhelming information is published on the Web, new information search and retrieval methods are needed in order to enable users to find more relevant information based not only on keywords, but also context and preferences of each individual user. This has lead to the introduction of a new era for Web information search and retrieval, namely, community-based and semantic-enhanced search.

The emergent Semantic Web technologies provide the possibility to realize the vision of meaningful relations and structured data on the Web. As an extension to the current Web, the Semantic Web technologies enable computers and people to work in cooperation (Berners-Lee, Hendler, & Lassila, 2001). The Semantic Web focuses on publishing and retrieving machine-processable Web contents (Dayal, Kuno, & Wilkinson, 2003). In the Semantic Web framework, flexible and interoperable structures such as Web ontology language (OWL)⁴ and resource description framework (RDF)⁵ are used to represent resources. The relationships between entities in a particular domain can be explicitly expressed using an ontology (Chandrasekaran, Josephson, & Benjamins, 1999). To describe multimedia data and documents on the Semantic Web, the ontology concepts are required to be mapped to the metadata description structure, which is usually referred to as semantic annotation. The semantic-enhanced search focuses on utilizing the structured description and knowledge description ontologies to enhance the results of information search and retrieval process on the Web. The better the relationships are processed and analyzed, the more relevant context results are obtained to be shown to users.