Automatic Turn Segmentation for Movie & TV Subtitles
2016 (English)In: 2016 IEEE Workshop on Spoken Language Technology, IEEE conference proceedings, 2016Conference paper (Refereed)
Movie and TV subtitles contain large amounts of conversational material, but lack an explicit turn structure. This paper present a data-driven approach to the segmentation of subtitles into dialogue turns. Training data is first extracted by aligning subtitles with transcripts in order to obtain speaker labels. This data is then used to build a classifier whose task is to determine whether two consecutive sentences are part of the same dialogue turn. The approach relies on linguistic, visual and timing features extracted from the subtitles themselves and does not require access to the audiovisual material -- although speaker diarization can be exploited when audio data is available. The approach also exploits alignments with related subtitles in other languages to further improve the classification performance. The classifier achieves an accuracy of 78% on a held-out test set. A follow-up annotation experiment demonstrates that this task is also difficult for human annotators.
Place, publisher, year, edition, pages
IEEE conference proceedings, 2016.
Computer Science Language Technology (Computational Linguistics)
IdentifiersURN: urn:nbn:se:kth:diva-193938OAI: oai:DiVA.org:kth-193938DiVA: diva2:1034694
2016 IEEE Workshop on Spoken Language Technology
QC 201610142016-10-122016-10-122016-10-14Bibliographically approved