topline-hook-scorer-da-ridge

Danish Hook Quality Ridge Baseline

A lightweight scikit-learn Ridge regression baseline for ranking Danish lyric lines.

Important caveat

This is a weakly supervised heuristic distillation model. The target is derived from TopLine Core features (ordlyd, singability, rhyme reuse, line length, section hints, stream normalization). It has not been trained on human accept/reject labels yet.

Use it as a first ranking signal for exploration, not as a final artistic judge.

Data

• Source dataset: Larslouvre/topline-core-corpus
• Danish songs: 247
• Lines: 3062
• Train/test split: grouped by song to avoid same-song leakage

Metrics

{
  "test_mae": 0.03193977245295291,
  "test_rmse": 0.03905283644419383,
  "test_r2": 0.8126409065765499,
  "test_spearman": 0.9022890369452443,
  "test_ndcg_at_20": 0.9593355974531179,
  "cv_mae_mean": 0.03411350540973291,
  "cv_mae_std": 0.0014743542672338325,
  "cv_rmse_mean": 0.040963221908164225,
  "cv_rmse_std": 0.0016030566426361575,
  "cv_r2_mean": 0.7070031835299886,
  "cv_r2_std": 0.01658186566627316,
  "n_lines": 3062,
  "n_songs": 247,
  "n_train_lines": 2145,
  "n_test_lines": 917,
  "n_train_songs": 197,
  "n_test_songs": 50,
  "target_mean": 0.581674631776957,
  "target_std": 0.08216859661908116,
  "trained_at": "2026-05-09T02:31:09.646660",
  "target_type": "weak_heuristic_distillation"
}

Usage

import joblib
import pandas as pd

model = joblib.load('model.joblib')
X = pd.DataFrame([{
    'line_text': 'jeg ser månen blive blokeret af dig',
    # include numeric feature columns from feature_columns.json
}])
score = model.predict(X)[0]

<!-- ml-intern-provenance -->

Generated by ML Intern

This model repository was generated by ML Intern, an agent for machine learning research and development on the Hugging Face Hub.

• Try ML Intern: https://smolagents-ml-intern.hf.space
• Source code: https://github.com/huggingface/ml-intern