This work describes an approach for the semi-automatic learning object metadata markup of course?s web pages and their posterior extraction into SCORM Packages. This is identified as the first step to recycle course?s web pages with the full potential of the semantic web. The key issue of this approach is to avoid as much as possible manual metadata markup. To achieve this goal, the process of automatic metadata markup is provided with: (i) an ontology of course descriptions in OWL that will basically provide a sound specification of the diverse elements in a course and (ii) context information explicitly set up by linguistics rules. A prototype implementation has been developed in Java for Spanish course?s web pages.