"""Lightweight section heading classification.
Classifies heading text into academic paper section categories
using keyword matching only. No position-based guessing.
"""
from __future__ import annotations
from .models import SectionSpan, CONFIDENCE_FALLBACK
# Category keywords mapped to labels, ordered by weight (highest first).
# When multiple keywords match, highest-weighted category wins.
CATEGORY_KEYWORDS: list[tuple[str, list[str], float]] = [
("results", ["result", "findings", "outcomes"], 1.0),
("conclusion", ["conclusion", "concluding"], 1.0),
("methods", ["method", "materials", "experimental", "procedure",
"protocol", "design", "participants", "subjects"], 0.85),
("abstract", ["abstract"], 0.75),
("background", ["background", "literature review", "related work"], 0.7),
("discussion", ["discussion"], 0.65),
("introduction", ["introduction"], 0.5),
("appendix", ["appendix", "supplementa", "acknowledgment", "acknowledgement",
"grant", "funding", "disclosure", "conflict of interest"], 0.3),
("references", ["reference", "bibliography"], 0.1),
]
# "summary" is special — matches conclusion unless combined with data words.
SUMMARY_EXCLUDES = ["statistics", "table", "data", "results summary"]
def categorize_heading(heading: str) -> tuple[str | None, float]:
"""Determine category from heading text using keyword matching.
Returns (category, weight) or (None, 0) if no match.
"""
heading_lower = heading.lower()
for category, keywords, weight in CATEGORY_KEYWORDS:
for keyword in keywords:
if keyword in heading_lower:
return (category, weight)
# Special handling for "summary" (only if no other category matched above).
# Exclusion-based: an unrecognised keyword just falls through to conclusion,
# rather than requiring a specific keyword for the base functionality.
if "summary" in heading_lower:
for exclude in SUMMARY_EXCLUDES:
if exclude in heading_lower:
return ("results", 1.0)
return ("conclusion", 1.0)
return (None, 0.0)
def assign_section(char_start: int, spans: list[SectionSpan]) -> str:
"""Find the section label for a given character position."""
label, _ = assign_section_with_confidence(char_start, spans)
return label
def assign_section_with_confidence(
char_start: int, spans: list[SectionSpan]
) -> tuple[str, float]:
"""Find section label and confidence for a character position."""
for span in spans:
if span.char_start <= char_start < span.char_end:
return span.label, span.confidence
return "unknown", CONFIDENCE_FALLBACK
