candidates_df = detect_body_headings(line_df, threshold=3.0)

candidates_df[["text", "heading_score", "candidate_level",
               "signal_font_size_ratio", "signal_has_numeric_prefix"]].head()