SEGMENTING OF RECORDED LECTURE VIDEOS - The Algorithm VoiceSeg
Stephan Repp, Christoph Meinel · 2006
In the past decade, we have witnessed a dramatic increase in the availability of online academic lecture videos. There are technical problems in the use of recorded lectures for learning: the problem of easy access to the multimedia lecture video content and the problem of finding the semantically appropriate information very quickly. The first step to a semantic lecture-browser is the segmenting of the large video-corpus into a smaller cohesion area. The task of breaking documents into topically coherent subparts is called topic segmentation. In this paper, we present a segmenting algorithm for recorded lecture videos based on their imperfect transcripts. The recorded lectures are transcripted by an out-of-the-box speech recognition software with a accuracy of approximately 70%-80%. Words as well as a time stamp for each word are stored in a database. This data acts as the input to our algorithm. We will show that the clustering of similar words, the generation of vectors with the values from the clusters and the calculation of the cosine-mass of adjacent vectors, leads to a better segmenting result compared to a standard algorithm.