File size: 14,167 Bytes
2e511b5 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 | """Goldenset data-quality audit: empty vs. valid vs. malformed per variable.
This does *not* judge whether a label is the right answer (there is no
reference) — it checks whether each filled cell conforms to the variable's
expected type/format, so we can report coding hygiene. Three buckets per cell:
- ``empty`` — None / blank (the value is simply absent),
- ``valid`` — filled and conforms to the variable's type/format/range/vocab,
- ``invalid`` — filled but malformed, e.g. free text where a number is
expected, a ratio outside ``[0, 1]``, a non-ISO date, or an ISIC code that is
not in the controlled vocabulary.
Outputs two tables (CSV + Markdown), matching the request:
1. **By country** — one row per (country, variable) with absolute counts.
2. **Aggregated by variable** — one row per variable, summed over all
countries, with counts and percentages.
Usage:
uv run legex-goldenset-quality
uv run legex-goldenset-quality --countries ch,de,br --out data/analysis/quality
"""
import argparse
import csv
import logging
import re
import sys
from collections import defaultdict
from datetime import date, datetime
from pathlib import Path
from openpyxl import load_workbook
from legex.config import settings
from legex.evaluation import is_label_column, normalise
from legex.utils import countries_with_goldenset, goldenset_path, goldenset_sheet
log = logging.getLogger("legex.goldenset_quality")
# --- Field type groups (the eleven substantive variables) -------------------
DATE_FIELDS = frozenset({"trial_start_date", "trial_end_date"})
RATIO_FIELDS = frozenset({"plaintiff_loosing_share"})
MONEY_FIELDS = frozenset(
{"court_cost_awarded_nominal", "party_compensation_awarded_nominal"}
)
# dispute_value_nominal is money but also allows the literal "nonpecuniary".
DISPUTE_FIELD = "dispute_value_nominal"
COUNT_FIELDS = frozenset({"plaintiffs_all_count", "defendants_all_count"})
ISIC_FIELDS = frozenset(
{"plaintiff_no1_ISIC1_industry_category", "defendant_no1_ISIC1_industry_category"}
)
STRING_FIELDS = frozenset({"legal_subject_judgement"})
# The eleven substantive schema variables. Stray columns (e.g. a workbook that
# duplicates the identifier as "Case Id") are ignored so the report stays on
# the schema.
SCHEMA_FIELDS = (
DATE_FIELDS | RATIO_FIELDS | MONEY_FIELDS | {DISPUTE_FIELD}
| COUNT_FIELDS | ISIC_FIELDS | STRING_FIELDS
)
# Controlled ISIC vocabulary (v3 prompt) plus the two documented fallbacks.
ISIC_VOCAB = frozenset(
{
"a_agriculture_forestry_fishing", "b_mining_quarrying", "c_manufacturing",
"d_electricity_gas_steam_ac", "e_water_sewerage_waste_remediation",
"f_construction", "g_wholesale_retail_trade", "h_transportation_storage",
"i_accommodation_food_service", "j_publishing_broadcasting_content",
"k_telecom_it_info_services", "l_financial_insurance", "m_real_estate",
"n_professional_scientific_technical", "o_administrative_support",
"p_public_admin_defence", "q_education", "r_human_health_social_work",
"s_arts_entertainment_recreation", "t_other_service_activities",
"u_households_as_employers", "v_extraterritorial_organisations",
"no_allocation_possible",
}
)
BUCKETS = ("empty", "valid", "invalid")
_NUM_RE = re.compile(r"^-?\d+(?:\.\d+)?$")
def _is_number(s: str) -> bool:
"""A clean plain number per the codebook (period decimal, no separators)."""
return bool(_NUM_RE.match(s.strip()))
def _is_int(s: str) -> bool:
s = s.strip()
if _NUM_RE.match(s):
f = float(s)
return f.is_integer() and f >= 0
return False
def _is_iso_date(value: object, normalised: str) -> bool:
if isinstance(value, (date, datetime)):
return True
try:
date.fromisoformat(normalised.strip())
return True
except ValueError:
return False
def classify_cell(field: str, raw: object) -> str:
"""Return one of BUCKETS for a single (field, raw cell value)."""
s = normalise(raw)
if not s:
return "empty"
if field in DATE_FIELDS:
return "valid" if _is_iso_date(raw, s) else "invalid"
if field == DISPUTE_FIELD:
if s.lower() == "nonpecuniary":
return "valid"
return "valid" if _is_number(s) else "invalid"
if field in MONEY_FIELDS:
return "valid" if _is_number(s) else "invalid"
if field in RATIO_FIELDS:
if not _is_number(s):
return "invalid"
return "valid" if 0.0 <= float(s) <= 1.0 else "invalid"
if field in COUNT_FIELDS:
return "valid" if _is_int(s) else "invalid"
if field in ISIC_FIELDS:
return "valid" if s.lower() in ISIC_VOCAB else "invalid"
if field in STRING_FIELDS:
# Free text: a bare number where a translated legal subject is expected
# is the malformed case; otherwise any non-empty string is acceptable.
return "invalid" if _is_number(s) else "valid"
# Unknown / currency-like columns are excluded upstream; default to valid.
return "valid"
# (country, field) -> {bucket: count}
Counts = dict[tuple[str, str], dict[str, int]]
def audit_country(cc: str) -> tuple[dict[str, dict[str, int]], int]:
"""Return ({field: {bucket: count}}, n_rows) for one country."""
path = goldenset_path(cc)
wb = load_workbook(path, read_only=True, data_only=True)
try:
ws = goldenset_sheet(wb)
rows = ws.iter_rows(values_only=True)
header = [str(c) if c is not None else "" for c in next(rows)]
label_cols = [h for h in header if is_label_column(h) and h in SCHEMA_FIELDS]
per_field: dict[str, dict[str, int]] = {
f: {b: 0 for b in BUCKETS} for f in label_cols
}
n_rows = 0
for row in rows:
cells = dict(zip(header, row))
# Count only substantively-annotated rows
if not any(cells.get(f) not in (None, "") for f in label_cols):
continue
n_rows += 1
for f in label_cols:
per_field[f][classify_cell(f, cells.get(f))] += 1
return per_field, n_rows
finally:
wb.close()
def _expected_hint(field: str) -> str:
"""Human-readable description of the valid form, for the hand-cleaning worklist."""
if field in DATE_FIELDS:
return "ISO date YYYY-MM-DD"
if field == DISPUTE_FIELD:
return "number or 'nonpecuniary'"
if field in MONEY_FIELDS:
return "number (period decimal, no thousands separators / currency symbols)"
if field in RATIO_FIELDS:
return "number in [0, 1]"
if field in COUNT_FIELDS:
return "integer >= 0"
if field in ISIC_FIELDS:
return "ISIC category from the controlled vocab, or no_allocation_possible"
if field in STRING_FIELDS:
return "text (not a bare number)"
return ""
def collect_invalid(cc: str) -> list[tuple[str, str, str, str, str]]:
"""Return one (country, case_id, field, raw_value, expected) row per invalid cell."""
path = goldenset_path(cc)
wb = load_workbook(path, read_only=True, data_only=True)
out: list[tuple[str, str, str, str, str]] = []
try:
ws = goldenset_sheet(wb)
rows = ws.iter_rows(values_only=True)
header = [str(c) if c is not None else "" for c in next(rows)]
label_cols = [h for h in header if is_label_column(h) and h in SCHEMA_FIELDS]
for row in rows:
cells = dict(zip(header, row))
if not any(cells.get(f) not in (None, "") for f in label_cols):
continue
case_id = cells.get("case_id")
case_id = str(case_id) if case_id not in (None, "") else ""
for f in label_cols:
raw = cells.get(f)
if classify_cell(f, raw) == "invalid":
out.append((cc, case_id, f, "" if raw is None else str(raw), _expected_hint(f)))
return out
finally:
wb.close()
def _pct(n: int, total: int) -> float:
return n / total if total else 0.0
def write_by_country_csv(counts: Counts, path: Path) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
with path.open("w", encoding="utf-8", newline="") as f:
w = csv.writer(f)
w.writerow(["country", "variable", "n", "empty", "valid", "invalid",
"pct_empty", "pct_valid", "pct_invalid"])
for (cc, field), c in sorted(counts.items()):
total = c["empty"] + c["valid"] + c["invalid"]
w.writerow([cc, field, total, c["empty"], c["valid"], c["invalid"],
f"{_pct(c['empty'], total):.4f}",
f"{_pct(c['valid'], total):.4f}",
f"{_pct(c['invalid'], total):.4f}"])
def write_by_variable_csv(agg: dict[str, dict[str, int]], path: Path) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
with path.open("w", encoding="utf-8", newline="") as f:
w = csv.writer(f)
w.writerow(["variable", "n", "empty", "valid", "invalid",
"pct_empty", "pct_valid", "pct_invalid"])
for field, c in sorted(agg.items()):
total = c["empty"] + c["valid"] + c["invalid"]
w.writerow([field, total, c["empty"], c["valid"], c["invalid"],
f"{_pct(c['empty'], total):.4f}",
f"{_pct(c['valid'], total):.4f}",
f"{_pct(c['invalid'], total):.4f}"])
def write_invalid_csv(rows: list[tuple[str, str, str, str, str]], path: Path) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
with path.open("w", encoding="utf-8", newline="") as f:
w = csv.writer(f)
w.writerow(["country", "case_id", "field", "raw_value", "expected"])
w.writerows(rows)
def render_markdown(counts: Counts, agg: dict[str, dict[str, int]]) -> str:
lines = ["# Goldenset data-quality audit", ""]
lines.append("Buckets: **empty** (absent), **valid** (well-typed), "
"**invalid** (filled but malformed, e.g. text where a number "
"is expected, ratio outside [0,1], non-ISO date, unknown ISIC).")
lines.append("")
lines.append("## Aggregated by variable (all countries)")
lines.append("")
lines.append("| Variable | n | empty | valid | invalid | %empty | %valid | %invalid |")
lines.append("|---|---:|---:|---:|---:|---:|---:|---:|")
for field, c in sorted(agg.items()):
total = c["empty"] + c["valid"] + c["invalid"]
lines.append(
f"| `{field}` | {total} | {c['empty']} | {c['valid']} | {c['invalid']} | "
f"{_pct(c['empty'], total):.1%} | {_pct(c['valid'], total):.1%} | "
f"{_pct(c['invalid'], total):.1%} |"
)
lines.append("")
lines.append("## By country (absolute counts)")
lines.append("")
lines.append("| Country | Variable | n | empty | valid | invalid |")
lines.append("|---|---|---:|---:|---:|---:|")
for (cc, field), c in sorted(counts.items()):
total = c["empty"] + c["valid"] + c["invalid"]
lines.append(
f"| `{cc}` | `{field}` | {total} | {c['empty']} | {c['valid']} | {c['invalid']} |"
)
return "\n".join(lines)
def main(argv: list[str] | None = None) -> int:
logging.basicConfig(level=logging.INFO, format="%(message)s")
parser = argparse.ArgumentParser(description="Goldenset data-quality audit.")
parser.add_argument("--countries", default=None, help="Comma-separated codes (default: all).")
parser.add_argument("--out", type=Path, default=None,
help="Output dir (default data/analysis/quality).")
parser.add_argument("--list-invalid", action="store_true",
help="Also write invalid_cells.csv: one row per invalid cell "
"(country, case_id, field, raw_value, expected) for hand-cleaning.")
args = parser.parse_args(argv)
countries = (
[c.strip() for c in args.countries.split(",") if c.strip()]
if args.countries else countries_with_goldenset()
)
out_dir = args.out or (settings.data_dir / "analysis" / "quality")
counts: Counts = {}
agg: dict[str, dict[str, int]] = defaultdict(lambda: {b: 0 for b in BUCKETS})
for cc in countries:
gs = goldenset_path(cc)
if not gs.exists():
log.warning("[%s] no goldenset, skipping", cc)
continue
per_field, n_rows = audit_country(cc)
log.info("[%s] %d rows, %d variables", cc, n_rows, len(per_field))
for field, c in per_field.items():
counts[(cc, field)] = c
for b in BUCKETS:
agg[field][b] += c[b]
if not counts:
log.error("no goldensets scored")
return 1
write_by_country_csv(counts, out_dir / "by_country.csv")
write_by_variable_csv(agg, out_dir / "by_variable.csv")
report = render_markdown(counts, dict(agg))
(out_dir / "report.md").write_text(report, encoding="utf-8")
if args.list_invalid:
invalid_rows: list[tuple[str, str, str, str, str]] = []
for cc in countries:
if goldenset_path(cc).exists():
invalid_rows.extend(collect_invalid(cc))
write_invalid_csv(invalid_rows, out_dir / "invalid_cells.csv")
log.info("wrote %s (%d invalid cells)",
out_dir / "invalid_cells.csv", len(invalid_rows))
# Console summary: the aggregated-by-variable table.
width = max((len(f) for f in agg), default=len("variable"))
print(f"\n{'variable'.ljust(width)} {'n':>5} {'empty':>6} {'valid':>6} {'invalid':>7}")
for field, c in sorted(agg.items()):
total = c["empty"] + c["valid"] + c["invalid"]
print(f"{field.ljust(width)} {total:>5} "
f"{_pct(c['empty'], total):>6.1%} {_pct(c['valid'], total):>6.1%} "
f"{_pct(c['invalid'], total):>7.1%}")
log.info("\nwrote %s, %s, %s",
out_dir / "by_country.csv", out_dir / "by_variable.csv", out_dir / "report.md")
return 0
if __name__ == "__main__":
sys.exit(main())
|