Spaces:
Sleeping
Sleeping
File size: 13,523 Bytes
9381502 87b5875 9381502 87b5875 9381502 87b5875 9381502 87b5875 9381502 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 | from __future__ import annotations
import json
import re
from collections import Counter, defaultdict
from pathlib import Path
from statistics import median
from typing import Any
from cert_study_app.services.question_type_metadata_service import is_visual_question_type, normalize_question_type
ISSUE_WEIGHTS = {
"missing_stem": 18,
"short_stem": 8,
"missing_answer": 10,
"missing_options": 8,
"option_label_gap": 8,
"answer_not_in_options": 8,
"duplicate_raw_text": 10,
"number_gap": 12,
"number_duplicate": 12,
"page_regression": 10,
"chunk_too_short": 8,
"chunk_too_long": 6,
"answer_leaked_to_stem": 8,
"embedded_next_question": 14,
"image_missing": 6,
}
def default_quality_report_path(output_json: str | Path) -> str:
path = Path(output_json)
return path.with_name(f"{path.stem}.quality.json").as_posix()
def load_parsed_questions(json_path: str | Path) -> list[dict[str, Any]]:
with Path(json_path).open("r", encoding="utf-8") as f:
data = json.load(f)
if isinstance(data, dict) and isinstance(data.get("questions"), list):
data = data["questions"]
if not isinstance(data, list):
raise ValueError("parsed JSON must be a list or an object with a questions list")
return [item for item in data if isinstance(item, dict)]
def build_parse_quality_report(
json_path: str | Path,
*,
output_path: str | Path | None = None,
expected_count: int | None = None,
) -> dict[str, Any]:
questions = load_parsed_questions(json_path)
issues_by_index: dict[int, list[dict[str, Any]]] = defaultdict(list)
numbers = [_as_int(_first_present(item, "number", "question_number")) for item in questions]
pages = [_as_int(item.get("page")) for item in questions]
raw_fingerprints: dict[str, list[int]] = defaultdict(list)
chunk_lengths = []
for index, item in enumerate(questions):
raw_text = _text(item.get("raw_text") or item.get("ocr_text") or item.get("stem") or item.get("question"))
stem = _text(item.get("stem") or item.get("question") or item.get("q_text"))
answer = item.get("answer")
options = _normal_options(item.get("options"))
question_type = normalize_question_type(item.get("question_type") or "mcq")
raw_fingerprints[_fingerprint(raw_text)].append(index)
chunk_lengths.append(len(raw_text))
if not stem:
_add_issue(issues_by_index, index, "missing_stem", "๋ฌธ์ ๋ณธ๋ฌธ์ด ๋น์ด ์์ต๋๋ค.")
elif len(stem) < 30:
_add_issue(issues_by_index, index, "short_stem", "๋ฌธ์ ๋ณธ๋ฌธ์ด ๋๋ฌด ์งง์ต๋๋ค.", length=len(stem))
if not _has_answer(answer):
_add_issue(issues_by_index, index, "missing_answer", "์ ๋ต์ด ๋น์ด ์์ต๋๋ค.")
if not is_visual_question_type(question_type) and len(options) < 2:
_add_issue(issues_by_index, index, "missing_options", "๊ฐ๊ด์/์ ํํ์ผ๋ก ๋ณด์ด์ง๋ง ๋ณด๊ธฐ๊ฐ 2๊ฐ ๋ฏธ๋ง์
๋๋ค.", option_count=len(options))
if len(options) >= 2:
expected_labels = [chr(ord("A") + offset) for offset in range(len(options))]
labels = [label for label, _body in options]
if labels != expected_labels:
_add_issue(
issues_by_index,
index,
"option_label_gap",
"๋ณด๊ธฐ ๋ผ๋ฒจ์ด A๋ถํฐ ์ฐ์๋์ง ์์ต๋๋ค.",
labels=labels,
expected=expected_labels,
)
if _answer_labels(answer) and not set(_answer_labels(answer)).issubset(set(labels)):
_add_issue(
issues_by_index,
index,
"answer_not_in_options",
"์ ๋ต ๋ผ๋ฒจ์ด ๋ณด๊ธฐ ๋ผ๋ฒจ ์์ ์์ต๋๋ค.",
answer_labels=_answer_labels(answer),
option_labels=labels,
)
if re.search(r"(?im)^\s*(?:Answer|์ ๋ต)\s*:", stem):
_add_issue(issues_by_index, index, "answer_leaked_to_stem", "๋ณธ๋ฌธ์ ์ ๋ต ๋ผ์ธ์ด ์์ฌ ์์ต๋๋ค.")
if re.search(r"(?m)\n\s*\d{1,3}\s*[.)]\s+\S+", stem):
_add_issue(issues_by_index, index, "embedded_next_question", "์ฒญํฌ ์์ ๋ค์ ๋ฌธ์ ์์์ฒ๋ผ ๋ณด์ด๋ ์ค์ด ์์ต๋๋ค.")
if is_visual_question_type(question_type) and item.get("image_path") and not Path(str(item.get("image_path"))).exists():
_add_issue(issues_by_index, index, "image_missing", "์ด๋ฏธ์ง ๊ธฐ๋ฐ ๋ฌธ์ ์ ์๋ฌธ ์ด๋ฏธ์ง ํ์ผ์ ์ฐพ์ ์ ์์ต๋๋ค.")
nonzero_numbers = [number for number in numbers if number is not None]
duplicates = [number for number, count in Counter(nonzero_numbers).items() if count > 1]
if duplicates:
for index, number in enumerate(numbers):
if number in duplicates:
_add_issue(issues_by_index, index, "number_duplicate", "๋ฌธ์ ๋ฒํธ๊ฐ ์ค๋ณต๋์์ต๋๋ค.", number=number)
gaps = _number_gaps(nonzero_numbers)
if gaps:
gap_numbers = set()
for start, end in gaps:
gap_numbers.update({start - 1, end + 1})
for index, number in enumerate(numbers):
if number in gap_numbers:
_add_issue(issues_by_index, index, "number_gap", "๋ฌธ์ ๋ฒํธ ์ฐ์์ฑ์ด ๊นจ์ง ์ง์ ๊ทผ์ฒ์
๋๋ค.", gaps=gaps[:10])
previous_page = None
for index, page in enumerate(pages):
if page is None:
continue
if previous_page is not None and page < previous_page:
_add_issue(issues_by_index, index, "page_regression", "ํ์ด์ง ๋ฒํธ๊ฐ ์ด์ ๋ฌธํญ๋ณด๋ค ์์ต๋๋ค.", previous_page=previous_page, page=page)
previous_page = page
repeated_fingerprints = {fp: indexes for fp, indexes in raw_fingerprints.items() if fp and len(indexes) > 1}
for indexes in repeated_fingerprints.values():
for index in indexes:
_add_issue(issues_by_index, index, "duplicate_raw_text", "๊ฐ์ ์๋ฌธ ์ฒญํฌ๊ฐ ์ค๋ณต ํ์ฑ๋์์ต๋๋ค.", duplicate_indexes=indexes)
length_stats = _length_stats(chunk_lengths)
short_cutoff = 50
long_cutoff = max(2500, int((length_stats.get("median") or 0) * 4))
for index, length in enumerate(chunk_lengths):
if length and length < short_cutoff:
_add_issue(issues_by_index, index, "chunk_too_short", "์๋ฌธ ์ฒญํฌ๊ฐ ๋๋ฌด ์งง์ต๋๋ค.", length=length)
elif long_cutoff and length > long_cutoff:
_add_issue(issues_by_index, index, "chunk_too_long", "์๋ฌธ ์ฒญํฌ๊ฐ ๋น์ ์์ ์ผ๋ก ๊น๋๋ค.", length=length, cutoff=long_cutoff)
issue_counts = Counter(issue["code"] for issues in issues_by_index.values() for issue in issues)
total_penalty = sum(ISSUE_WEIGHTS.get(code, 5) * count for code, count in issue_counts.items())
count_penalty = 0
if expected_count is not None and expected_count >= 0:
count_delta = abs(len(questions) - expected_count)
count_penalty = min(20, count_delta * 2)
score = max(0, min(100, 100 - total_penalty - count_penalty))
samples = []
for index, issues in sorted(issues_by_index.items(), key=lambda row: (-_issue_weight_sum(row[1]), row[0]))[:30]:
item = questions[index]
samples.append(
{
"index": index,
"number": _first_present(item, "number", "question_number"),
"page": item.get("page"),
"question_type": item.get("question_type"),
"issues": issues,
"stem_preview": _preview(item.get("stem") or item.get("question") or ""),
"raw_preview": _preview(item.get("raw_text") or item.get("ocr_text") or ""),
}
)
report = {
"schema_version": 1,
"source_json": Path(json_path).as_posix(),
"score": score,
"status": _status_for_score(score, issue_counts),
"question_count": len(questions),
"expected_count": expected_count,
"issue_counts": dict(sorted(issue_counts.items())),
"question_issues": [
{
"index": index,
"number": _first_present(questions[index], "number", "question_number"),
"page": questions[index].get("page"),
"issues": issues,
}
for index, issues in sorted(issues_by_index.items())
],
"metrics": {
"missing_answer": issue_counts.get("missing_answer", 0),
"missing_options": issue_counts.get("missing_options", 0),
"needs_review_estimate": len(issues_by_index),
"numbers": {
"first": min(nonzero_numbers) if nonzero_numbers else None,
"last": max(nonzero_numbers) if nonzero_numbers else None,
"duplicates": duplicates[:30],
"gaps": gaps[:30],
},
"pages": {
"first": next((page for page in pages if page is not None), None),
"last": next((page for page in reversed(pages) if page is not None), None),
"regressions": issue_counts.get("page_regression", 0),
},
"chunk_lengths": length_stats,
},
"samples": samples,
}
if output_path:
Path(output_path).parent.mkdir(parents=True, exist_ok=True)
Path(output_path).write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8")
return report
def summarize_quality_report(report: dict[str, Any]) -> str:
counts = report.get("issue_counts") or {}
top = ", ".join(f"{key} {value}" for key, value in sorted(counts.items(), key=lambda row: (-row[1], row[0]))[:4])
return f"ํ์ง ์ ์ {report.get('score', 0)}์ ยท {report.get('question_count', 0)}๋ฌธํญ" + (f" ยท {top}" if top else "")
def _add_issue(target: dict[int, list[dict[str, Any]]], index: int, code: str, message: str, **details: Any) -> None:
target[index].append({"code": code, "message": message, "severity": _severity(code), "details": details})
def _severity(code: str) -> str:
weight = ISSUE_WEIGHTS.get(code, 5)
if weight >= 12:
return "high"
if weight >= 8:
return "medium"
return "low"
def _issue_weight_sum(issues: list[dict[str, Any]]) -> int:
return sum(ISSUE_WEIGHTS.get(str(issue.get("code")), 5) for issue in issues)
def _text(value: Any) -> str:
return str(value or "").strip()
def _first_present(item: dict[str, Any], *keys: str) -> Any:
for key in keys:
if item.get(key) not in {None, ""}:
return item.get(key)
return None
def _as_int(value: Any) -> int | None:
try:
return int(value)
except Exception:
return None
def _normal_options(raw: Any) -> list[tuple[str, str]]:
if isinstance(raw, dict):
rows = [(str(key).strip().upper(), str(value).strip()) for key, value in raw.items()]
elif isinstance(raw, list):
rows = []
for index, value in enumerate(raw):
text = str(value or "").strip()
match = re.match(r"^([A-Za-z])[\.\)]\s*(.+)$", text, re.S)
if match:
rows.append((match.group(1).upper(), match.group(2).strip()))
else:
rows.append((chr(ord("A") + index), text))
else:
rows = []
return [(label, body) for label, body in rows if label or body]
def _has_answer(value: Any) -> bool:
if isinstance(value, list):
return bool(value)
if isinstance(value, dict):
return bool(value)
text = str(value or "").strip()
return bool(text and text not in {"[]", "{}"})
def _answer_labels(value: Any) -> list[str]:
if isinstance(value, list):
text = " ".join(str(item) for item in value)
elif isinstance(value, dict):
text = " ".join(str(item) for item in value.values())
else:
text = str(value or "")
text = text.upper()
if re.fullmatch(r"[A-H]{2,8}", text.strip()):
return list(text.strip())
labels = []
for token in re.findall(r"\b[A-H]\b|\b[1-8]\b", text):
labels.append(chr(ord("A") + int(token) - 1) if token.isdigit() else token)
return labels
def _fingerprint(text: str) -> str:
normalized = re.sub(r"\s+", " ", text or "").strip().lower()
if len(normalized) < 80:
return ""
return normalized[:500]
def _number_gaps(numbers: list[int]) -> list[tuple[int, int]]:
if len(numbers) < 2:
return []
unique_sorted = sorted(set(numbers))
gaps = []
for previous, current in zip(unique_sorted, unique_sorted[1:]):
if current > previous + 1:
gaps.append((previous + 1, current - 1))
return gaps
def _length_stats(lengths: list[int]) -> dict[str, Any]:
if not lengths:
return {"min": 0, "median": 0, "max": 0}
return {"min": min(lengths), "median": int(median(lengths)), "max": max(lengths)}
def _preview(value: Any, limit: int = 220) -> str:
text = re.sub(r"\s+", " ", str(value or "")).strip()
return text[:limit]
def _status_for_score(score: int, issue_counts: Counter) -> str:
if issue_counts.get("number_gap") or issue_counts.get("embedded_next_question"):
return "needs_chunk_review"
if score >= 85:
return "pass"
if score >= 65:
return "needs_sampling"
return "needs_reparse"
|