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

LinkedTV at MediaEval 2014 search and hyperlinking task

Le, Hoang An; Bui, Q.M; Huet, Benoit; Cervenková, B; Bouchner, J; Apostolidis, E.; Markatopoulou, F; Pournaras, A; Mezaris, V; Stein, D; Eickeler, S; Stadtschnitzer, M

MEDIAEVAL 2014, MediaEval Benchmarking Initiative for Multimedia Evaluation Workshop, October 16-17, 2014, Barcelona, Spain

The paper presents the LinkedTV approaches for the Search and Hyperlinking (S&H) task at MediaEval 2014. Our submissions aim at evaluating 2 key dimensions: temporal gran- ularity and visual properties of the video segments. The temporal granularity of target video segments is de ned by grouping text sentences, or consecutive automatically de- tected shots, considering the temporal coherence, the visual similarity and the lexical cohesion among them. Visual properties are combined with text search results using mul- timodal fusion for re-ranking. Two alternative methods are proposed to identify which visual concepts are relevant to each query: using WordNet similarity or Google Image analysis. For Hyperlinking, relevant visual concepts are identi ed by analysing the video anchor.

Document Bibtex

Titre:LinkedTV at MediaEval 2014 search and hyperlinking task
Département:Data Science
Eurecom ref:4429
Copyright: CEUR
Bibtex: @inproceedings{EURECOM+4429, year = {2014}, title = {{L}inked{TV} at {M}edia{E}val 2014 search and hyperlinking task}, author = {{L}e, {H}oang {A}n and {B}ui, {Q}.{M} and {H}uet, {B}enoit and {C}ervenkov{\'a}, {B} and {B}ouchner, {J} and {A}postolidis, {E}. and {M}arkatopoulou, {F} and {P}ournaras, {A} and {M}ezaris, {V} and {S}tein, {D} and {E}ickeler, {S} and {S}tadtschnitzer, {M}}, booktitle = {{MEDIAEVAL} 2014, {M}edia{E}val {B}enchmarking {I}nitiative for {M}ultimedia {E}valuation {W}orkshop, {O}ctober 16-17, 2014, {B}arcelona, {S}pain}, address = {{B}arcelona, {ESPAGNE}}, month = {10}, url = {} }
Voir aussi: