Skip to Main content Skip to Navigation
Conference papers

Segmentation et Regroupement en Locuteur pour le traitement incrémental des collections volumineuses

Abstract : Current cross-show diarization systems are mainly based on an overall clustering process that handles collectively each show of a collection. This approach has already been studied in various situations and seems to be the best way so far to achieve low error rates. Nevertheless, that process shows its limits in a realistic applicative context where large and dynamically increasing collections have to be processed. In this paper we investigate the use of an incremental clustering cross-show speaker diarization architecture to iteratively process new shows within an existing collection. The new shows to be inserted are processed one after another, according to the chronological order of broadcasting. Experiments were conducted on the LCP and the BFMTV show recordings distributed among the ETAPE and the REPERE French evaluation campaigns. It represents 67 hours of annotated data, distributed among 310 shows, and covering a 2-years period (from Sept. 2010 to Oct. 2012).
Document type :
Conference papers
Complete list of metadata

Cited literature [16 references]  Display  Hide  Download
Contributor : sylvain meignier Connect in order to contact the contributor
Submitted on : Friday, April 7, 2017 - 9:12:01 AM
Last modification on : Tuesday, December 8, 2020 - 9:44:18 AM
Long-term archiving on: : Saturday, July 8, 2017 - 12:27:47 PM


Publisher files allowed on an open archive


  • HAL Id : hal-01433245, version 1



Grégor Dupuy, Sylvain Meignier, yannick Estève. Segmentation et Regroupement en Locuteur pour le traitement incrémental des collections volumineuses. 30e Journées d’Études sur la Parole (JEP'14), 2014, Le Mans, France. pp.433 - 440. ⟨hal-01433245⟩



Record views


Files downloads