Ecole d'ingénieur et centre de recherche en Sciences du numérique

Detecting hot spots in Web videos

Redondo Garcia, José Luis; Sabatino, Mariela; Lisena, Pasquale; Troncy, Raphaël

ISWC 2014, 13th International Semantic Web Conference, Demo Track, October 21-23, 2014, Riva del Garda, Italy / CEUR Proceedings, Vol. 1272

This paper presents a system that detects and enables the exploration of relevant fragments (called Hot Spots) inside educational online videos. Our approach combines visual analysis techniques and background knowledge from the web of data in order to quickly get an overview about the video content and therefore promote media consumption at the fragment level. First, we perform a chapter segmentation by combining visual features and semantic units (paragraphs) available in transcripts. Second, we semantically annotate those segments via Named Entity Extraction and topic detection. We then identify consecutive segments talking about similar topics and entities that we merge into bigger and semantic independent media units. Finally, we rank those segments and filter out the lowest scored candidates, in order to propose a summary that illustrates the Hot Spots in a dedicated media player. An online demo is available at

Document Doi Bibtex

Titre:Detecting hot spots in Web videos
Mots Clés:Semantic Video Annotation, Media Fragments, Summarization
Type:Poster / Demo
Ville:Riva del Garda
Département:Data Science
Eurecom ref:4399
Copyright: CEUR
Bibtex: @poster / demo{EURECOM+4399, year = {2014}, title = {{D}etecting hot spots in {W}eb videos}, author = {{R}edondo {G}arcia, {J}os{\'e} {L}uis and {S}abatino, {M}ariela and {L}isena, {P}asquale and {T}roncy, {R}apha{\"e}l}, number = {EURECOM+4399}, month = {10}, institution = {Eurecom} address = {{R}iva del {G}arda, {ITALIE}}, url = {} }
Voir aussi: