In this paper, we describe the system we used to address the Hyperlinking task at TRECVID 2016 and the achieved results. Our system combines several features (textual and visual) in order to consider the different facets of the input videos. Specifically, we combined automatically generated transcripts, visual concepts, and the text extracted by means of an OCR tool. We also exploited WordNet to find synonyms which are used to apply a simple query expansion technique. The four submitted runs aimed at analyzing the impact of the considered features on the quality of the retrieved hyperlinks.
Type:
Conférence
City:
Gaithersburg
Date:
2016-11-14
Department:
Data Science
Eurecom Ref:
5163
Copyright:
© NIST. Personal use of this material is permitted. The definitive version of this paper was published in and is available at :
See also:
