OCR-aided person annotation and label propagation for speaker modeling in TV shows - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

OCR-aided person annotation and label propagation for speaker modeling in TV shows

Résumé

In this paper, we present an approach for minimizing human effort in manual speaker annotation. Label propagation is used at each iteration of an active learning cycle. More precisely, a selection strategy for choosing the most suitable speech track to be labeled is proposed. Four different selection strategies are evaluated and all the tracks in a corresponding cluster are gathered using agglomerative clustering in order to propagate human annotations. To further reduce the manual labor required, an optical character recognition system is used to bootstrap annotations. At each step of the cycle , annotations are used to build speaker models. The quality of the generated speaker models is evaluated at each step using an i-vector based speaker identification system. The presented approach shows promising results on the REPERE corpus with a minimum amount of human effort for annotation .
Fichier principal
Vignette du fichier
ocr-aided-person-4.pdf (168.89 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01350071 , version 1 (29-07-2016)

Identifiants

Citer

Mateusz Budnik, Laurent Besacier, Ali Khodabakhsh, Cenk Demiroglu. OCR-aided person annotation and label propagation for speaker modeling in TV shows. IEEE ICASSP 2016, Mar 2016, Shangai, China. ⟨10.1109/ICASSP.2016.7472743⟩. ⟨hal-01350071⟩
145 Consultations
131 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More