Semantic structuring of video collections fromspeech : segmentation and hyperlinking

Abstract : New challenges emerged in the past years as the audiovisual landscape significantly transformed with the emergence of Internet-based TV. The decision of what to watch and in what order no longer belongs to the TV station but the user. New solutions must be offered to users in order to facilitate precise and quick access to the desired information contained in the ever increasing amounts of audiovisual content: for example, to identify a specific event, a TV show fragment containing a certain public person or addressing a particular topic, or to browse sequentially across all segments of videos addressing this topic. These new and varied features imply the development of innovative methods for structuring and exploiting the audiovisual content, which represent the focus of this thesis. We offer automatic topic structuring techniques of audiovisual data to reveal the internal organization of each program. We also study the implications of the structure produced on various tasks such as creating hyperlinks between videos (to allow navigation between TV show fragments that are thematically related) or creating automatic summaries. All the work is carried out on automatic transcripts of the speech pronounced in TV shows, to achieve generic solutions, not dedicated to a particular kind of show.
Complete list of metadatas

Cited literature [144 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01253678
Contributor : Abes Star <>
Submitted on : Wednesday, March 9, 2016 - 11:31:34 AM
Last modification on : Friday, January 11, 2019 - 3:15:16 PM
Long-term archiving on: Monday, June 13, 2016 - 9:33:29 AM

File

SIMON_Anca_Roxana.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01253678, version 2

Citation

Anca-Roxana Şimon. Semantic structuring of video collections fromspeech : segmentation and hyperlinking. Formal Languages and Automata Theory [cs.FL]. Université Rennes 1, 2015. English. ⟨NNT : 2015REN1S076⟩. ⟨tel-01253678v2⟩

Share

Metrics

Record views

302

Files downloads

441