We propose a method for extractive summarization of audiovisual recordings focusing on topic-level segments. We first build a content similarity graph between all segments across the collection, using word vectors from the transcripts, and then select the ...
ACM Press2016