Evaluation of spoken language recognition technology using broadcast speech: performance and challenges.

摘要: Spoken Language Recognition (SLR) technology has remarkably improved in the last years, partly thanks to NIST Evaluations (LRE), which have become standard benchmarks for testing new approaches. evaluations focus on narrow-band conversational telephone speech and deal with some specific target languages. Recent efforts expand scope of SLR assessment include Albayzin 2008 2010 LRE, wide-band TV broadcast speech. In this work, a system based state-of-the-art approaches is developed evaluated LRE datasets, looking identify those conditions that make task challenging eventually guide design future using same kind data. We present analyse performance under different conditions, regarding: (1) set languages (including details about confusion each other) amount data available estimate models; (3) presence background noise.