Semantic heterogeneity is becoming increasingly prominent in bioinformatics domains that deal with constantly expanding, dynamic, often very large, datasets from various distributed sources. Metadata is the key component for effective information integration. Traditional approaches for reconciling semantic heterogeneity use standards or mediation-based methods. These approaches have had limited success in addressing the general semantic heterogeneity problem and by themselves are not likely to succeed in bioinformatics domains where one faces the additional complexity of keeping pace with the speed at which data and semantic heterogeneity is being generated. This paper presents a methodology for reconciliation of semantic heterogeneity of metadata in bioinformatics data sources. The approach is based on the proposition that by globally monitoring, clustering, and visualizing bioinformatics metadata across disparately created data sources, patterns of practice can be identified. This can facilitate semantic reconciliation of metadata in current data and mitigate semantic heterogeneity in future data by promoting sharing and reuse of existing metadata. To instantiate the methodology, a research architecture, MicroSEEDS, is presented and its implementation and envisioned uses are discussed.

The problem of semantic heterogeneity has gained ever increasing attention as the world's digital information systems become more closely interconnected. Among the various interoperability problems faced by information integration solutions, the semantic interoperability problem has been growing in importance, with increasingly sophisticated approaches being proposed for its resolution. This article reviews the essential characteristics of the semantic heterogeneity problem and its solution approaches in various domains, and proposes frameworks for evaluating the problem and selecting an appropriate solution approach.