ModernBERT Disfluency Detection — Exp A (Real Only)

Fine-tuned from answerdotai/ModernBERT-base on real-only FluencyBank Timestamped data.

Dataset

  • Config: real_only de arielcerdap/disfluency-fluencybank
  • Train: 2737 segmentos (100% reales)
  • Val/Test: idénticos a Exp B para comparación directa

Labels

O · FP (filled pause) · RP (repetition) · RV (revision) · PW (partial word)

Test Results

Label P R F1 Support
Label P R F1 Support
--- --- --- --- ---
O 0.9783 0.9860 0.9821 3704
FP 0.9888 1.0000 0.9944 176
RP 0.7784 0.8276 0.8022 174
RV 0.3425 0.2907 0.3145 86
PW 0.9510 0.8326 0.8879 233

Macro F1 (4 disfluencias): 0.7497
Binary F1: 0.8902

Hyperparameters

  • learning_rate: 5e-05
  • epochs: 15
  • warmup_steps: 191
  • weight_decay: 0.1
  • focal_loss_gamma: 3.0 (adaptive)
  • class_weights: O=1.0, FP=3.0, RP=6.0, RV=20.0, PW=5.0
Downloads last month
9
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train arielcerdap/modernbert-disfluency-expA-realonly