Content-based resource description is the key to find appropriate information sources that are most likely to contain the relevant documents for a given user query. However, semantic heterogeneity makes it difficult to acquire accurate and meaningful resource descriptions from distributed, heterogeneous information sources. To address this problem, we describe an ontology-based approach which uses domain-specific ontologies to extract content-related information from information sources, and to generate ontology-based resource descriptions. The preliminary experimental results demonstrate that our ontology-based approach could improve selection accuracy.