With the widespread use of Information Technology for teaching support, it is usual to made digital content available through Digital Library Systems (DLMS) or Learning Management Systems (LMS).These systems, however, work independently, have different characteristics and manipulate different types of materials, and their data and metadata repositories are heterogeneous and distributed. The content of repositories would be better used if it was integrated in the same environment or accessed in an integrated way from DLMS and LMS. This thesis presents a homogeneous view of DLMS and LMS content. In order to provide such homogenization, it is proposed an extension of the mediator and wrapper architecture for dealing with metadata integration and ontologies for treating semantics. Local ontologies are used for describing each metadata repository, and a global ontology for the integration. As documents of DLMS repositories tend to be monolithic and not to follow a reuse approach, rules for extracting the most important content from these documents were developed in order to make them reusable. This extraction includes text mining techniques as well as rules for discovering definitions embedded in the documents. A prototype was developed which implements the extraction and proves the feasibility of this approach. In order to make the work easier to understand, it is presented a case study that uses the proposed technique and the prototype. The work described in this thesis facilitates and enriches the development of learning material by making the content of digital library documents reusable and integrated to existing learning objects.