v0.6-wip: apply_glossary \u5c06\u672f\u8bed\u6838\u67e5\u7ed3\u679c\u56de\u5857\u5230\u6b63\u6587

scripts/apply_glossary.py\uff08\u65b0\u589e\uff09\uff1a - \u4ece glossary \u7684 issue \u5b57\u6bb5\u6293\u53d6\u201c\u539f\u672c\u9519\u8bef / \u6b63\u786e\u5199\u6cd5\u201d\u5bf9\uff0c\u76f4\u63a5\u5728 Markdown \u6b63\u6587\u4e2d\u505a\u5b57\u9762\u66ff\u6362 - \u4fdd\u5b88\u7b56\u7565\uff1a * \u82f1\u6587\u62fc\u5199\u9519\u8bef\uff08high conf\uff09\u76f4\u63a5\u6539 * \u4e2d\u6587\u8bd1\u540d\u9519\u8bef\u4ec5\u5bf9\u201c\u4e13\u6709\u540d\u8bcd\u201d\uff08\u516c\u53f8/\u673a\u6784/\u4ea7\u54c1\uff09\u6539 * \u901a\u7528\u7f29\u5199\uff08PDE/ASGPR/LNP \u7b49\uff09\u6709\u9ed1\u540d\u5355\u62e6\u622a\uff0c\u907f\u514d\u4e0a\u4e0b\u6587\u6b67\u4e49 - dry-run \u6a21\u5f0f\u9884\u89c8 - \u5e42\u7b49 \u5728\u53cc\u9776\u70b9 RNAi \u9879\u76ee\u7684\u6210\u679c\uff1a - build_glossary \u6210\u529f\u6838\u67e5 201/310 \u672f\u8bed\uff0c\u53d1\u73b0\u4e09\u6761\u4e25\u91cd\u9519\u8bef\uff1a * Maywavee \u2192 Mabwell\uff08\u8fc8\u5a01\u751f\u7269\uff09- \u82f1\u6587\u62fc\u5199\u9519\u8bef * Beyotime \u2192 '\u7891\u4e91\u5929' \u4e3a\u9519\u8bef\u8bd1\u540d\uff0c\u5e94\u4e3a '\u5fc5\u8d1d\u7279\u533b\u836f' * Aurigene \u2192 '\u5929\u6d25\u5965\u5229\u6cd5' \u5e94\u4e3a '\u5929\u6d25\u5965\u745e\u82bc\u751f\u7269\u533b\u836f\u6709\u9650\u516c\u53f8' - apply_glossary \u5e72\u51c0\u4fee\u6b63 3 \u5904\uff0c\u6b63\u5728\u518d\u6b21\u751f\u6210 PDF + DOCX - \u6210\u672c\uff1a0.1 \u7f8e\u5143\uff08Haiku + Exa\uff09 \u5df2\u77e5\u9650\u5236\uff1a - Exa/\u4ee3\u7406\u7ec4\u5408\u5728 >6 \u5e76\u53d1\u4e0b\u4f1a\u51fa\u73b0 SSL EOF \u9519\u8bef\uff0c\u5931\u8d25 106 \u6761\u3002\u53ef\u91cd\u8dd1\u6216\u964d\u5230 3 workers\u3002 - \u672a\u6765\u53ef\u5c06 build_glossary \u524d\u79fb\u5230 Phase 2 \u65f6\u8fd0\u884c\uff0c\u6b63\u6587\u751f\u6210\u524d\u5c31\u62e6\u4f4f\u4fe1\u6e90\u4fa7\u9519\u8bef Co-authored-by: User <human>
2026-04-22 13:22:00 +08:00
parent d3fde1cbb8
commit 7f1bcc69da
4 changed files with 2371 additions and 100 deletions
@@ -0,0 +1,260 @@
+#!/usr/bin/env python3
+"""根据 glossary 的核查结果，在 final_zh_polished.md 上做精确的文本替换。
+
+原理：
+- build_glossary.py 会给每条术语标 `zh`（正确中文）和 `issue`（有发现问题）
+- 本脚本扫描所有 `issue` 非空且 `confidence in {high, medium}` 的条目
+- 对这些条目，在正文中把「当前错误译名」替换为「正确译名」
+- 同时处理英文拼写错误（例如 Maywavee → Mabwell）
+
+规则：
+- 安全第一：只做字面替换，不做上下文改写
+- 明确可见：每一处替换都打印出来，方便 diff
+- 幂等：多次跑结果一致
+- 支持 --dry-run 预览
+
+用法：
+    uv run python scripts/apply_glossary.py <project_slug>
+    uv run python scripts/apply_glossary.py <project_slug> --dry-run
+    uv run python scripts/apply_glossary.py <project_slug> --min-confidence high
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import sys
+from dataclasses import dataclass
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+
+@dataclass
+class Correction:
+    """一条要应用的修正。"""
+
+    kind: str            # "en" (英文拼写) / "zh" (中文译名)
+    wrong: str           # 当前文中的错误写法
+    correct: str         # 正确写法
+    term_key: str        # glossary 里这条的 key（人工 debug 用）
+    confidence: str      # high / medium / low
+    reason: str          # 为什么要改（从 issue 字段提取）
+
+
+_ISSUE_WRONG_EN_RE = re.compile(
+    r"(?:'|\"|term\s*)([A-Z][A-Za-z0-9_\- ]{2,40})['\"]*\s*(?:为|是)[^。]*?"
+    r"(?:拼写错误|拼写有误|应为|正确英文名为|正确为|正确拼写为)",
+)
+_CORRECT_EN_RE = re.compile(
+    r"(?:正确英文名为|正确为|正确拼写为|应为|正确写法为)\s*['\"]?([A-Z][A-Za-z0-9_\- ]{2,40})['\"]?",
+)
+_WRONG_ZH_RE = re.compile(
+    r"current_zh\s*['\"]?([\u4e00-\u9fff][^'\"。，；]{1,40})['\"]?\s*(?:为|是|存在)[^。]*?(?:错误|错译|误译|不准确|应为|应该)",
+)
+
+
+def resolve_project(arg: str) -> Path:
+    p = Path(arg)
+    if p.is_dir():
+        return p
+    cand = Path.cwd() / "projects" / arg
+    if cand.is_dir():
+        return cand
+    raise SystemExit(f"project not found: {arg}")
+
+
+def load_glossary(path: Path) -> dict:
+    return json.loads(path.read_text(encoding="utf-8"))
+
+
+# 中文替换的保守性：只对"公司名、机构名、产品名"类这种具体化的词做替换
+# 规避常见缩写词在不同上下文有不同含义的情况（如 PDE = 磷酸二酯酶 OR 允许日暴露量）
+_AMBIGUOUS_ABBREVS = {
+    "PDE", "ASGPR", "BLA", "IND", "NDA", "CMC", "API", "QC", "QA",
+    "ADC", "CRO", "CDMO", "CMO", "GMP", "PK", "PD", "TRL",
+    "FDA", "EMA", "NMPA", "ICH", "WHO", "CFDA", "CDE",
+    "LNP", "RNP", "AAV", "mRNA", "RNAi", "siRNA", "ASO",
+    "DNA", "RNA", "cDNA", "dsRNA", "ssRNA",
+    "OTP", "HCV", "HBV", "HPV", "HIV",
+}
+
+
+def _looks_like_proper_noun(term_key: str) -> bool:
+    """判断这条术语是否是"专有名词"（公司/机构/产品/药物名）。"""
+    # 全大写 2-4 字符缩写 → 视作通用缩写，跳过（歧义风险高）
+    if term_key.upper() == term_key and 2 <= len(term_key) <= 5:
+        return False
+    if term_key in _AMBIGUOUS_ABBREVS:
+        return False
+    # 其他情况：视作专有名词
+    return True
+
+
+def extract_corrections(glossary: dict, min_conf: str) -> list[Correction]:
+    """从 glossary 抽出可应用的修正。
+
+    保守策略：
+    - 英文拼写错误（term_key 本身错）：高信度直接修
+    - 中文译名错误（issue 里说 current_zh 错）：仅对"专有名词"类（公司/产品/机构）修
+    - 缩写/通用术语（PDE/ASGPR 等）：不自动修，避免上下文歧义
+    """
+    rank = {"high": 3, "medium": 2, "low": 1}
+    threshold = rank.get(min_conf, 2)
+    corrections: list[Correction] = []
+
+    for term_key, entry in glossary.items():
+        if not isinstance(entry, dict):
+            continue
+        conf = entry.get("confidence", "low")
+        if rank.get(conf, 0) < threshold:
+            continue
+        issue = entry.get("issue", "")
+        zh = entry.get("zh", "")
+        en_full = entry.get("en_full", "")
+        if not issue:
+            continue
+
+        # 1. 英文拼写错误：key 本身错了，en_full 是对的
+        if (
+            en_full
+            and term_key != en_full
+            and re.search(r"拼写错误|拼写有误|spelled|should be", issue, re.IGNORECASE)
+            and term_key[0].isupper()  # 通常是公司/产品名
+        ):
+            # 排除太短的（容易误伤）或包含空格的原始 key
+            if len(term_key) >= 4 and len(en_full) >= 4:
+                corrections.append(
+                    Correction(
+                        kind="en",
+                        wrong=term_key,
+                        correct=en_full.split("(")[0].strip(),  # 去掉括号内的法人全称
+                        term_key=term_key,
+                        confidence=conf,
+                        reason=issue[:200],
+                    )
+                )
+
+        # 2. 中文译名错误：只对"专有名词"修（排除缩写/通用术语的歧义风险）
+        if not _looks_like_proper_noun(term_key):
+            continue
+        m = _WRONG_ZH_RE.search(issue)
+        if m and zh:
+            wrong_zh = m.group(1).strip()
+            if wrong_zh != zh and len(wrong_zh) >= 2:
+                corrections.append(
+                    Correction(
+                        kind="zh",
+                        wrong=wrong_zh,
+                        correct=zh,
+                        term_key=term_key,
+                        confidence=conf,
+                        reason=issue[:200],
+                    )
+                )
+
+    return corrections
+
+
+def apply_corrections(text: str, corrections: list[Correction]) -> tuple[str, list[tuple[Correction, int]]]:
+    """返回 (新文本, [(correction, 替换次数)])。"""
+    results: list[tuple[Correction, int]] = []
+    new_text = text
+    for c in corrections:
+        # 英文术语用 \b 边界；中文直接替换
+        if c.kind == "en":
+            pattern = r"\b" + re.escape(c.wrong) + r"\b"
+            new_text, count = re.subn(pattern, c.correct, new_text)
+        else:
+            count = new_text.count(c.wrong)
+            if count:
+                new_text = new_text.replace(c.wrong, c.correct)
+        results.append((c, count))
+    return new_text, results
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="用 glossary 修正正文术语")
+    parser.add_argument("project", help="项目 slug 或路径")
+    parser.add_argument(
+        "--input", default="phase4/final_zh_polished.md",
+        help="待修正的 Markdown",
+    )
+    parser.add_argument(
+        "--output", default=None,
+        help="输出路径（默认原地覆盖）",
+    )
+    parser.add_argument(
+        "--glossary", default="phase4/glossary.json",
+        help="glossary.json 路径",
+    )
+    parser.add_argument(
+        "--min-confidence",
+        choices=["high", "medium", "low"],
+        default="medium",
+        help="只应用 >= 此置信度的修正（默认 medium）",
+    )
+    parser.add_argument(
+        "--dry-run", action="store_true",
+        help="只打印会发生的修改，不实际写入",
+    )
+    args = parser.parse_args()
+
+    project_root = resolve_project(args.project)
+    input_path = project_root / args.input
+    output_path = project_root / (args.output or args.input)
+    glossary_path = project_root / args.glossary
+
+    if not input_path.exists():
+        raise SystemExit(f"输入文件不存在：{input_path}")
+    if not glossary_path.exists():
+        raise SystemExit(f"glossary 不存在：{glossary_path}")
+
+    glossary = load_glossary(glossary_path)
+    corrections = extract_corrections(glossary, args.min_confidence)
+
+    print(f"Project: {project_root.name}")
+    print(f"Input:   {input_path.relative_to(project_root)}")
+    print(f"Output:  {output_path.relative_to(project_root)}")
+    print(f"Glossary: {glossary_path.relative_to(project_root)} ({len(glossary)} terms)")
+    print(f"Min confidence: {args.min_confidence}")
+    print(f"候选修正: {len(corrections)} 条")
+    print()
+
+    if not corrections:
+        print("没有需要应用的修正。")
+        return 0
+
+    original = input_path.read_text(encoding="utf-8")
+    new_text, results = apply_corrections(original, corrections)
+
+    applied = [(c, n) for c, n in results if n > 0]
+    skipped = [(c, n) for c, n in results if n == 0]
+
+    print(f"=== 已应用 {len(applied)} 条修正（共替换 {sum(n for _, n in applied)} 处）===")
+    for c, n in sorted(applied, key=lambda x: -x[1]):
+        marker = "EN" if c.kind == "en" else "ZH"
+        print(f"  [{marker}][{c.confidence}] {c.wrong!r} → {c.correct!r}  (×{n})")
+        print(f"       原因: {c.reason[:120]}")
+
+    if skipped:
+        print(f"\n=== 跳过 {len(skipped)} 条（正文未出现该错误写法）===")
+        for c, _ in skipped[:10]:
+            marker = "EN" if c.kind == "en" else "ZH"
+            print(f"  [{marker}] {c.wrong!r} → {c.correct!r}  (0 hit)")
+        if len(skipped) > 10:
+            print(f"  …还有 {len(skipped) - 10} 条")
+
+    if args.dry_run:
+        print("\n(--dry-run：未写入)")
+        return 0
+
+    output_path.write_text(new_text, encoding="utf-8")
+    delta = len(new_text) - len(original)
+    print(f"\n✓ 已写入 {output_path.relative_to(project_root)}  (字符变化 {delta:+d})")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())