Semantic indexing for recorded educational lecture videos

In this paper, we present a general architecture and a new retrieval method for an educational system that is based on a knowledge base of existing recorded lectures. The extraction of metadata from the multimedia resources is one of the main parts of this paper. The recorded lectures are transcripted by an out-of-the-box speech recognition software. The speech recognition software generates a time stamp for each word. These resources are divided into clusters, marked by timestamps, so that search engines are able to find the exact position of particular information inside a course. Finally, a retrieval method is presented that allows users to find "example", "explanation", "overview", "repetition", "exercise" for a particular word or topic-word. This is useful for the student's learning process and allows easy navigation of the multimedia database for pervasive learning.