ASR

The 'ASR' dataset/benchmark is used to evaluate automatic speech recognition performance, particularly in the context of non-native speech and the effectiveness of discrete speech tokens derived from self-supervised learning models.

Papers using ASR (8)

A Comparative Study On E-branchformer Vs Conformer In Speech Recognition, Translation, And Understanding Tasks2023 · 11 citesDiscrete Speech Unit Extraction via Independent Component Analysis2025 · 4 citesWhisperner: Unified Open Named Entity And Speech Recognition2024 · 1 citesLeveraging Soft Distributions of SSL-Derived Discrete Speech Tokens for Downstream Inference2026Steer-MoE: Efficient Audio-Language Alignment with a Mixture-of-Experts Steering Module2025Entropy-based Coarse and Compressed Semantic Speech Representation Learning2025WhisperNER: Unified Open Named Entity and Speech Recognition2024 · 1 citesCoBERT: Self-Supervised Speech Representation Learning Through Code Representation Learning2022