This model predicts grammar scores (1–5) from audio files. It uses Whisper to transcribe speech to text, cleans the text, and extracts features with TF-IDF. A Random Forest Regressor is trained to learn grammar score patterns. Evaluation via Pearson Correlation showed good results.
machine-learning random-forest speech-recognition tf-idf nlp-machine-learning model-evaluation pearson-correlation text-preprocessing regression-model audio-to-text whisper-model grammar-scoring submission-pipeline
-
Updated
Jun 20, 2025 - Jupyter Notebook