MCPcopy Create free account
hub / github.com/OpenRaiser/PaperFlow / _looks_like_extraction_noise

Function _looks_like_extraction_noise

agents/reading-agent/main.py:4153–4170  ·  view source on GitHub ↗
(text: str)

Source from the content-addressed store, hash-verified

4151
4152def _looks_like_extraction_noise(text: str) -> bool:
4153 cleaned = _clean_text(text)
4154 if not cleaned:
4155 return False
4156 lowered = cleaned.lower()
4157 noise_markers = (
4158 "arxiv:",
4159 "[cs.",
4160 "correspondence to:",
4161 "copyright",
4162 "accepted by",
4163 "proceedings of",
4164 "doi:",
4165 )
4166 marker_hits = sum(1 for marker in noise_markers if marker in lowered)
4167 short_line_count = sum(1 for line in cleaned.splitlines() if 0 < len(line.strip()) <= 18)
4168 digit_ratio = sum(ch.isdigit() for ch in cleaned) / max(1, len(cleaned))
4169 return marker_hits >= 2 or short_line_count >= 6 or digit_ratio > 0.12
4170
4171
4172def _append_template_qa_block(
4173 lines: List[str],

Callers 1

generate_reading_reportFunction · 0.85

Calls 1

_clean_textFunction · 0.70

Tested by

no test coverage detected