🎙️ PathPlanner Wav2Vec2 Pronunciation & Speech Evidence Scorer

Fine-tuned multi-task Wav2Vec2 model for automated CEFR speech assessment (Pronunciation, Fluency, Prosody, Completeness).

Model Overview

Backbone Architecture: facebook/wav2vec2-base (94.6M parameters)
Target Dimensions: Multi-Task Regression (Pronunciation, Fluency, Prosody, Completeness)
Input Audio: 16 kHz Mono PCM WAV
Evaluation Benchmark: Pearson Correlation PCC = 0.7733, MAE = 0.0649