This paper is devoted to explore media correlation and media synchronization in a composite multimedia document, the so-called navigated hypermedia document in our language learning system, to facilitate the multimedia authoring, presentation, and access. Two levels of media correlation in temporal, spatial, and content domains are investigated: syntactic level correlation and semantic level correlation. We devise a capturing mechanism to record all the media streams and relations between them, including voice and event streams, for replaying the lecturing in a form as close as possible to the original classroom experience. The syntactic level correlation is based on specific timestamps of the media stream and used to reconstruct the recorded lecture for synchronized presentation. Furthermore, to integrate media objects with specific segments within the media stream, some computed synchronization processes are required to discover semantic content of the media. The proposed computed s...