ASLP-lab/FMSU-Bench

Existing benchmarks predominantly cater to macroscopic tasks and suffer from coarse annotation granularity. We construct FMSU-Bench, a pioneering Fine-grained Multi-dimensional Speech Understanding Benchmark.

• Scale & Scope: Comprises over 24,000 bilingual instances (Chinese/English), manually verified by domain experts.
• Comprehensive Taxonomy: Systematically covers 14 distinct speech dimensions structured into a 5-tier taxonomy:
  1. Speaker Demographics: Gender, Age, Accent
  2. Acoustic-Prosodic Features: Pitch, Speaking Rate, Rhythm, Voice Texture
  3. Affective and Semantic Reasoning: Emotion, Tone, Contextual Inference
  4. Acoustic Scene Analysis: Background Sound, Acoustic Environment
  5. Linguistic-Paralinguistic Integration: Paralinguistic Events, Transcription with Paralinguistic Tags (Evaluated via our novel PATA metric).