Metadata is an increasingly important aspect of resource discovery. Good metadata has the potential to increase discovery and reuse and to facilitate interoperability of digital assets. Using the domain of learning objects, issues associated with the application of metadata standards and the challenges in metadata creation are examined. A framework for customizable metadata authoring that addresses the issues and challenges is described. The framework consists of flexible metadata schema, metadata schema views, templates, collaborative metadata editing, contextual help, and an effective interface component selection. A prototype implementation of the framework, CLOMAT (Customizable Learning Objects Metadata Authoring Tool), is used to illustrate the framework in operation. An initial evaluation of this prototype indicates substantial productivity gains over conventional metadata creation tools.