SpeechBrain

FFASR Leaderboard Launched — a Benchmark for Evaluating ASR in Real Acoustic Conditions

Treble Technologies and Hugging Face have launched the FFASR Leaderboard, the first open community benchmark for evaluating automatic speech recognition (ASR) systems in far-field conditions. The benchmark shows that at low signal-to-noise ratio (SNR), far-field recognition error rates are several times higher than in near-field. The platform enables evaluation of both accuracy (word error rate, WER) and speed (real-time factor, RTFx) of models.

Treble Technologies Hugging Face OpenAI IBM Cohere Meta SpeechBrain