code / scripts /appendix_frequencies.py
anonymous
[code] Reproduction bundle.
2e511b5
Raw
History Blame Contribute Delete
6.75 kB
#!/usr/bin/env python3
"""Frequency material for the paper appendix: ISIC sectors and currencies.
Reads the 19 published goldensets (primary rows with a coded legal subject)
and writes
* data/analysis/figures/isic_frequencies.pdf — horizontal bar chart of the
ISIC top-level sectors observed for plaintiffs and defendants;
* data/analysis/tables/currency_frequencies.tex — frequency of the currency
labels over the three monetary fields.
Usage: uv run python scripts/appendix_frequencies.py
"""
import argparse
import sys
from collections import Counter
from pathlib import Path
REPO_ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(REPO_ROOT))
from legex import published # noqa: E402
from legex.analysis.countries import RELEASE_COUNTRIES # noqa: E402
ISIC_FIELDS = (
"plaintiff_no1_ISIC1_industry_category",
"defendant_no1_ISIC1_industry_category",
)
CURRENCY_FIELDS = (
"Currency_dispute_value_nominal",
"Currency_court_cost_awarded_nominal",
"Currency_party_compensation_awarded_nominal",
)
FIG_OUT = REPO_ROOT / "data/analysis/figures/isic_frequencies.pdf"
TAB_OUT = REPO_ROOT / "data/analysis/tables/currency_frequencies.tex"
def _norm(v) -> str:
return str(v).strip() if v is not None else ""
def collect(gold_dir: Path) -> tuple[Counter, Counter, Counter, int]:
isic_p, isic_d, currencies = Counter(), Counter(), Counter()
n_judgments = 0
for cc in RELEASE_COUNTRIES:
seen: set[str] = set()
for rec in published.iter_gold_rows(gold_dir, cc):
cid = _norm(rec.get("case_id"))
if not cid or cid in seen: # first row per case_id = primary
continue
seen.add(cid)
if not _norm(rec.get("legal_subject_judgement")):
continue # released but not fully coded
n_judgments += 1
p = _norm(rec.get(ISIC_FIELDS[0])).lower()
d = _norm(rec.get(ISIC_FIELDS[1])).lower()
if p:
isic_p[p] += 1
if d:
isic_d[d] += 1
for f in CURRENCY_FIELDS:
c = _norm(rec.get(f)).upper()
if c:
currencies[c] += 1
return isic_p, isic_d, currencies, n_judgments
def _pretty(sector: str) -> str:
# e.g. "m_real_estate" -> "M — Real estate"; fallback codes stay verbatim.
if sector in ("none", "no_allocation_possible"):
return sector.replace("_", r"\_")
if len(sector) > 2 and sector[1] == "_":
letter, rest = sector[0].upper(), sector[2:]
return f"{letter}{rest.replace('_', ' ').capitalize()}"
return sector.replace("_", " ")
def write_isic_chart(isic_p: Counter, isic_d: Counter) -> None:
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
sectors = sorted(set(isic_p) | set(isic_d), key=lambda s: -(isic_p[s] + isic_d[s]))
y = range(len(sectors))
p_vals = [isic_p[s] for s in sectors]
d_vals = [isic_d[s] for s in sectors]
fig, ax = plt.subplots(figsize=(7.0, 0.34 * len(sectors) + 1.2))
h = 0.38
bars_p = ax.barh(
[i - h / 2 for i in y], p_vals, height=h,
color="#1F4E79", edgecolor="black", linewidth=0.5, label="Plaintiff",
)
bars_d = ax.barh(
[i + h / 2 for i in y], d_vals, height=h,
color="#DD6E6E", edgecolor="black", linewidth=0.5, label="Defendant",
)
for bars in (bars_p, bars_d):
ax.bar_label(bars, fontsize=6.5, padding=2, color="#374151")
ax.set_yticks(list(y))
ax.set_yticklabels([_pretty(s).replace(r"\_", "_") for s in sectors], fontsize=8)
ax.invert_yaxis()
ax.set_xlabel("Judgments with the sector assigned", fontsize=9)
ax.set_xlim(0, max(max(p_vals), max(d_vals)) * 1.08)
ax.grid(axis="x", linestyle=":", linewidth=0.6, color="#9CA3AF", alpha=0.7)
ax.set_axisbelow(True)
ax.legend(title="Party", frameon=True, edgecolor="#6B7280", fontsize=9, title_fontsize=9)
ax.tick_params(axis="x", labelsize=8)
fig.tight_layout()
FIG_OUT.parent.mkdir(parents=True, exist_ok=True)
fig.savefig(FIG_OUT)
plt.close(fig)
print(f"wrote {FIG_OUT} ({len(sectors)} sectors)")
# The currency columns are auxiliary audit fields filled in free text; map the
# common synonyms to ISO codes, drop the explicit no-currency markers.
_CURRENCY_SYNONYMS = {"RUPEES": "INR", "EURO": "EUR", "EUROS": "EUR"}
_NO_CURRENCY = {"NONE", "N/A", "NA", "-"}
_TOP_N = 12
def write_currency_table(currencies: Counter) -> None:
cleaned: Counter = Counter()
for cur, n in currencies.items():
if cur in _NO_CURRENCY:
continue
cleaned[_CURRENCY_SYNONYMS.get(cur, cur)] += n
total = sum(cleaned.values())
top = cleaned.most_common(_TOP_N)
rest = cleaned.most_common()[_TOP_N:]
lines = [
"% Auto-generated by scripts/appendix_frequencies.py — do not edit by hand.",
r"\begin{table}[t]",
r"\caption{Currencies of the three monetary fields"
r" (\texttt{dispute\_value}, \texttt{court\_cost}, \texttt{party\_compensation})"
f" over all filled cells of the release with a recorded currency (n\\,=\\,{total:,})."
r" Labels as recorded by the annotators, with common synonyms mapped to ISO codes.}",
r"\label{tab:currencies}",
r"\vskip 0.05in",
r"\centering\small",
r"\begin{tabular}{@{}lrr@{}}",
r"\toprule",
r"\textbf{Currency} & \textbf{Cells} & \textbf{Share} \\",
r"\midrule",
]
for cur, n in top:
label = cur.title() if len(cur) > 3 else cur
lines.append(f"{label} & {n:,} & {100 * n / total:.1f}\\% \\\\")
if rest:
n_rest = sum(n for _, n in rest)
lines.append(
f"Other ({len(rest)} labels) & {n_rest:,} & {100 * n_rest / total:.1f}\\% \\\\"
)
lines += [r"\bottomrule", r"\end{tabular}", r"\end{table}", ""]
TAB_OUT.parent.mkdir(parents=True, exist_ok=True)
TAB_OUT.write_text("\n".join(lines), encoding="utf-8")
print(f"wrote {TAB_OUT} ({len(cleaned)} currencies after cleaning)")
def main() -> None:
ap = argparse.ArgumentParser(description=__doc__.splitlines()[0])
ap.add_argument("--gold-dir", type=Path, default=None,
help="published goldenset data directory (default: "
"submission/goldensets/data, else ../goldensets/data)")
args = ap.parse_args()
gold_dir = args.gold_dir or published.default_gold_dir(REPO_ROOT)
isic_p, isic_d, currencies, n = collect(gold_dir)
print(f"{n} annotated judgments scanned")
write_isic_chart(isic_p, isic_d)
write_currency_table(currencies)
if __name__ == "__main__":
main()