File size: 6,426 Bytes
796575b | 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 | #!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
FinEval 数据准备：CSV -> JSON
来源：HuggingFace SUFE-AIFLM-Lab/FinEval（cc-by-nc-sa-4.0）
对应 NAACL 2025 论文中的「金融学术知识」部分：4,661 道选择题 / 34 学科。
另外三部分（金融行业知识 1434 / 金融安全知识 1640 / 金融代理 616）官方未开源。
口径（与 C-Eval 基线对齐，保证两套基准可比）：
few-shot 取自 dev（每科恰好 5 题），评测集为 val；test 集官方不公开答案，不使用。
输出：data/fineval/json/<subject>/{dev,val}.json
字段与 C-Eval 导出保持一致：id / question / A / B / C / D / answer
"""
import csv
import glob
import json
import os
ROOT = r"D:\ceval-baseline"
SRC = os.path.join(ROOT, "data", "fineval", "extracted")
OUT = os.path.join(ROOT, "data", "fineval", "json")
# 34 学科中文名（用于 prompt 头部「以下是中国关于<学科>的单项选择题」）
CN = {
"accounting": "会计",
"advanced_financial_accounting": "高级财务会计",
"auditing": "审计",
"banking_practitioner_qualification_certificate": "银行从业资格",
"central_banking": "中央银行学",
"certified_management_accountant": "管理会计师",
"certified_practising_accountant": "注册会计师",
"china_actuary": "中国精算师",
"commercial_bank_finance": "商业银行金融",
"corporate_finance": "公司金融",
"corporate_strategy_and_risk_management": "公司战略与风险管理",
"cost_accounting": "成本会计",
"econometrics": "计量经济学",
"economic_law": "经济法",
"finance": "金融学",
"financial_engineering": "金融工程",
"financial_management": "财务管理",
"financial_markets": "金融市场",
"fund_qualification_certificate": "基金从业资格",
"futures_practitioner_qualification_certificate": "期货从业资格",
"insurance": "保险",
"intermediate_financial_accounting": "中级财务会计",
"international_economics": "国际经济学",
"international_finance": "国际金融",
"investments": "投资学",
"macroeconomics": "宏观经济学",
"management_accounting": "管理会计",
"microeconomics": "微观经济学",
"monetary_finance": "货币金融",
"political_economy": "政治经济学",
"public_finance": "财政学",
"securities_practitioner_qualification_certificate": "证券从业资格",
"statistics": "统计学",
"tax_law": "税法",
}
# FinEval 把 34 科分为金融/经济/会计/金融类考证 4 大类
GROUP = {
"finance": ["finance", "corporate_finance", "financial_engineering", "financial_management",
"financial_markets", "investments", "international_finance", "monetary_finance",
"commercial_bank_finance", "central_banking", "insurance"],
"economy": ["microeconomics", "macroeconomics", "econometrics", "international_economics",
"political_economy", "public_finance", "statistics", "economic_law"],
"accounting": ["accounting", "advanced_financial_accounting", "intermediate_financial_accounting",
"auditing", "cost_accounting", "management_accounting"],
"certificate": ["certified_practising_accountant", "certified_management_accountant",
"china_actuary", "banking_practitioner_qualification_certificate",
"securities_practitioner_qualification_certificate",
"fund_qualification_certificate",
"futures_practitioner_qualification_certificate",
"corporate_strategy_and_risk_management", "tax_law"],
}
SUBJ_GROUP = {s: g for g, ss in GROUP.items() for s in ss}
def load_split(split):
out = {}
for f in sorted(glob.glob(os.path.join(SRC, split, "*.csv"))):
subj = os.path.basename(f)[: -len(f"_{split}.csv")]
rows = []
with open(f, encoding="utf-8-sig") as fh:
for row in csv.DictReader(fh):
if row.get("answer") not in list("ABCD"):
continue
# 选项文本里可能残留 LaTeX 转义（如 3$\%$），统一清洗为普通字符
clean = lambda s: (s or "").replace("$\\%$", "%").replace("\\%", "%").strip()
rows.append({
"id": f"{subj}_{row['id']}",
"question": (row.get("question") or "").strip(),
"A": clean(row.get("A")),
"B": clean(row.get("B")),
"C": clean(row.get("C")),
"D": clean(row.get("D")),
"answer": row["answer"],
})
out[subj] = rows
return out
def main():
dev = load_split("dev")
val = load_split("val")
os.makedirs(OUT, exist_ok=True)
tot_dev = tot_val = 0
summary = []
for subj in sorted(dev):
if subj not in val:
print("!! val 缺少学科:", subj)
continue
d, v = dev[subj], val[subj]
os.makedirs(os.path.join(OUT, subj), exist_ok=True)
with open(os.path.join(OUT, subj, "dev.json"), "w", encoding="utf-8") as f:
json.dump(d, f, ensure_ascii=False, indent=1)
with open(os.path.join(OUT, subj, "val.json"), "w", encoding="utf-8") as f:
json.dump(v, f, ensure_ascii=False, indent=1)
tot_dev += len(d)
tot_val += len(v)
summary.append({
"subject": subj, "subject_cn": CN.get(subj, subj),
"group": SUBJ_GROUP.get(subj, "?"),
"n_dev": len(d), "n_val": len(v),
})
with open(os.path.join(ROOT, "data", "fineval", "inventory.json"), "w", encoding="utf-8") as f:
json.dump({"n_subjects": len(summary), "n_dev": tot_dev, "n_val": tot_val,
"subjects": summary}, f, ensure_ascii=False, indent=2)
print(f"学科数 {len(summary)} | dev {tot_dev} 题 | val {tot_val} 题")
# 缺中文名的学科
miss = [s["subject"] for s in summary if s["subject"] not in CN]
if miss:
print("!! 缺中文名:", miss)
# 分组统计
import collections
c = collections.Counter(s["group"] for s in summary)
print("按大类:", dict(c))
print("每科 dev 题数分布:", sorted({s["n_dev"] for s in summary}))
print("输出 ->", OUT)
if __name__ == "__main__":
main()
|