Approaches to multi-domain language recognition

M. K. NANDWANA; MITCHELL MCLAREN; LUCIANA FERRER; DIEGO CASTÁN

Les Sables D'Olonne

Congreso; Odyssey 2018; 2018

ISCA

Multi-domain language recognition involves the application of a language identification (LID) system to identify languages in more than one domain. This problem wasthe focus of the recent NIST LRE 2017, and this article presents the findings from the SRI team during sys-tem development for the evaluation. Approaches foundto provide robustness in multi-domain LID include adomain-and-language-weighted Gaussian backend classifier, duration-aware calibration, and a source normalized multi-resolution neural network backend. The recently developed speaker embeddings technology is alsoapplied to the task of language recognition, showing greatpotential for future LID research.