Nowadays there are lots of learning objects stored on numerous repositories. They are described by meta-data allowing users (either authors or learners) to search and reuse these learning objects. The process of reuse is not so simple and requires some attention during the design phase. In our opinion, it seems clear that semantic metadata are required to allow a real reusing and assembling of learning objects. Following this principle we have defined a system based on three semantic models used to describe the domain, learners and learning objects. In this paper we investigate how to improve the design process of resources described using our proposal. For this, we propose two kinds of approaches: metrics and types. Using metrics, authors may have information about several aspects of quality of their existing resources. Types allow constraining the authoring process by defining constraints to be enforced by new learning objects.