A uniform compact genetic algorithm for matching bibliographic ontologies

Abstract

Digital Library (DL) is a source of inspiration for the standards and technologies on Semantic Web (SW), which is usually implemented by using bibliographic data. To address DL’s data heterogeneity problem, it is necessary to annotate the bibliographic data with semantic information, which requires the utilization of the bibliographic ontologies. In particular, a bibliographic ontology provides the domain knowledge by specifying the domain concepts and their relationships. However, due to human subjectivity, a concept in different bibliographic ontologies might be defined in different names, causing the data heterogeneity problem. To address this issue, it is necessary to find the mappings between bibliographic ontologies’ concepts, which is the so-called bibliographic ontology matching. In this paper, a Uniform Compact Genetic Algorithm (UCGA) is proposed to match the bibliographic ontologies, which employs the real-valued compact encoding mechanism to improve the algorithm’s performance, the Linearly Decreasing Virtual Population (LDVP) to trade-off the exploration and exploitation of the algorithm, and the local perturbation to enhance the convergence speed and avoid the algorithm’s premature convergence. In addition, by using the Uniform Probability Density Function (UPDF) and Uniform Cumulative Distribution Function (UCDF), the UCGA can greatly reduce the evolutionary time and memory consumption. The experiment uses the Biblio testing cases provided by the Ontology Alignment Evaluation Initiative (OAEI) to evaluate UCGA’s performance and the experimental results show that UCGA is both effective and efficient.