duyunzhoudyz/dyz-gongwen-skills / skills/gongwen-speech
gongwen-speech
Verified起草讲话类材料:领导讲话(动员部署、推进、表彰、座谈、年度工作会、开学毕业典礼)、主持词、欢迎和开幕闭幕致辞、表态发言、交流发言、代表发言。按场合、讲话人、听众和时长(约200字/分钟)定结构字数,避开前后讲话重复,写成口语稿,引语只用用户给的原文。用户说写讲话稿、准备15分钟发言、开学典礼致辞、写主持词、表态发言、经验交流发言时使用。党课交 gongwen-party,大会工作报告和述职交 gongwen-report,发布会发布词交 gongwen-brief,会议通知交 gongwen-draft,已有讲话稿压缩交 gongwen-polish;装有行业包时,本行业专用文书交对应行业包。
scripts/speech_check.py
View on GitHub#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""讲话稿时长与口语化检查(gongwen-speech)。
按语速估算讲话时长,并找出不便口头表达的长句、数字过密和过长的段落,列出引语和来源不明的引文,
供起草者校核时长、整理引语核对清单。
计数口径:从称谓行(如“同志们:”)算到结尾,不含空白,含标点;标题、署名、日期行,
以及【待补:……】【待核实:……】标记内的文字不计。遇到单独成行的【起草说明】即停止读取。
用法:
python scripts/speech_check.py <讲话稿> [--minutes 分钟] [--rate 200] [--style 讲|念]
[--max-sentence N] [--max-numbers N] [--max-para N]
[--json] [--strict]
退出码:默认 0;--strict 时存在 P1 返回 1;输入错误返回 2。
只用标准库,不联网;读取 .docx 需要 python-docx。本脚本不写任何文件。
"""
from __future__ import annotations
import argparse
import json
import re
import sys
from pathlib import Path
from typing import Dict, List, Optional, Sequence, Tuple
VERSION = "1.0.0"
# 稿子用法对应的默认阈值(工具约定,见 references/oral-style.md)
STYLE_DEFAULTS: Dict[str, Dict[str, int]] = {
"讲": {"max_sentence": 40, "max_numbers": 3, "max_para": 250},
"念": {"max_sentence": 60, "max_numbers": 5, "max_para": 400},
}
DEFAULT_RATE = 200
HOUR_LIMIT_MINUTES = 60
# 称谓行最长字数:礼宾称谓可能点名多位领导和来宾,比标题区的行长
SALUTATION_MAX_CHARS = 150
TOLERANCE = 0.10
# 讲的稿超过这个字数仍只在开头称呼一次的,提示穿插称谓
SALUTATION_HINT_CHARS = 2000
# 【起草说明】分界行(兼容 Markdown 标题、加粗包裹的写法,与 gongwen_lint.py 一致)
DRAFT_NOTE_RE = re.compile(r"^\s*(?:#{1,6}\s*)?(?:(?:\*\*|__)\s*)?【起草说明】")
MARKER_RE = re.compile(r"【(待补|待核实)[::][^】]*】")
SENTENCE_END = "。!?!?;;"
SENTENCE_RE = re.compile(r"[^。!?!?;;]+[。!?!?;;]*")
HEADER_END_PUNCT = set("。!?!?;;")
TOP_HEADING_RE = re.compile(r"^\s*[一二三四五六七八九十]{1,3}、")
LIST_NUMBER_RE = re.compile(r"^\s*(?:\d{1,2}[..]|[((]\d{1,2}[))])")
# 数字:整数、小数、带千分位,带百分号或千分号;后接年、月、日的视为日期,不计。
# 日期范围(如“1—9月”“2026—2028年”)中前一个数也视为日期
NUMBER_RE = re.compile(r"\d+(?:[.,]\d+)*(?:\s*[%%‰])?")
DATE_TAIL_RE = re.compile(r"^\s*(?:[—–\--~~至到]\s*\d+\s*)?(年|月|日|号|时)")
# 引语:引号前紧挨着“指出、强调、说”等动词(可带冒号或逗号),
# 兼容“所指出的:”“强调的那样,”“重要论述:”“重要讲话:”等引法,
# 以及“强调,要“……””“指出,必须“……””这类动词后带逗号和连接词的部分引用
QUOTE_RE = re.compile(r"“([^”]{4,})”")
QUOTE_VERB_RE = re.compile(
r"(指出|强调|说|讲|讲道|讲到|写道|告诫|提出|提到|要求|号召|说过|曾说|曾经说过|论述|重要讲话|重要指示|批示|云|曰)"
r"(?:的(?:那样)?)?(?:[::]|[,,]\s*(?:要|必须|就是)?)?\s*$"
)
SOURCE_RE = re.compile(
r"《[^》]+》|在[^。!?“”]{2,40}(讲话|会议|会上|大会|文件|通知|意见|决定|方案|纪要|报告|信|批示|指示|座谈|考察|论述|回信|贺信)"
)
# 来源不明的引子:P1 为假托古人、名人;P2 为俗语和“正如……所说”(可能指同场讲话,需人工核对)
VAGUE_SOURCE_P1_RE = re.compile(r"(古人云|古语云|古人说|古话说|名人说|有句名言|一位哲人|一位伟人|伟人说)")
VAGUE_SOURCE_P2_RE = re.compile(r"(俗话说|常言道|有句话说|有句话叫|有句老话|正如[^,。!?“”]{1,15}所说)")
SALUTATION_WORDS_RE = re.compile(r"同志们|同学们|老师们|朋友们|各位同事|各位代表|各位来宾|各位居民")
class _ChineseHelpFormatter(argparse.RawDescriptionHelpFormatter):
"""把 argparse 的 usage 前缀改为中文。"""
def add_usage(self, usage, actions, groups, prefix=None): # type: ignore[override]
return super().add_usage(usage, actions, groups, "用法:" if prefix is None else prefix)
class InputError(Exception):
"""输入文件或参数有误。"""
def setup_console() -> None:
for stream in (sys.stdout, sys.stderr):
try:
stream.reconfigure(encoding="utf-8") # type: ignore[attr-defined]
except Exception: # 被重定向为 StringIO 等对象时没有 reconfigure
pass
def nonspace_len(text: str) -> int:
return sum(1 for ch in text if not ch.isspace())
def strip_markers(text: str) -> str:
return MARKER_RE.sub("", text)
def clip(text: str, limit: int = 30) -> str:
text = text.strip()
return text if len(text) <= limit else text[:limit] + "……"
# ---------------------------------------------------------------------------
# 读取
# ---------------------------------------------------------------------------
def read_lines(path: Path) -> List[Tuple[int, str]]:
"""读取文件,返回 [(行号或段号, 文本)]。"""
if not path.is_file():
raise InputError(f"未找到文件:{path}")
suffix = path.suffix.lower()
if suffix in (".txt", ".md", ".markdown", ".text"):
try:
raw = path.read_bytes()
except OSError as exc:
raise InputError(f"无法读取文件:{exc}") from exc
try:
text = raw.decode("utf-8-sig")
except UnicodeDecodeError:
try:
text = raw.decode("gb18030")
except UnicodeDecodeError as exc:
raise InputError("无法识别文件编码,请另存为 UTF-8 后重试。") from exc
text = text.replace("\r\n", "\n").replace("\r", "\n")
return list(enumerate(text.split("\n"), start=1))
if suffix == ".docx":
try:
import docx # type: ignore
except ImportError as exc:
raise InputError("读取 .docx 需要 python-docx,请先运行:pip install python-docx") from exc
try:
document = docx.Document(str(path))
except Exception as exc: # 损坏或加密的 docx
raise InputError(f"无法打开 docx 文件:{exc}") from exc
return [(i, p.text) for i, p in enumerate(document.paragraphs, start=1)]
raise InputError(f"不支持的文件类型:{path.suffix or '无扩展名'}。只支持 .txt、.md、.docx。")
# ---------------------------------------------------------------------------
# 结构识别
# ---------------------------------------------------------------------------
class Speech:
"""讲话稿正文:识别标题区、称谓行,按段落保存。"""
def __init__(self, lines: Sequence[Tuple[int, str]]) -> None:
self.truncated_line: Optional[int] = None
paras: List[Tuple[int, str]] = []
for number, text in lines:
if DRAFT_NOTE_RE.match(text):
self.truncated_line = number
break
if text.strip():
paras.append((number, text.strip()))
self.paras = paras
self.header: List[Tuple[int, str]] = []
self.salutation: Optional[Tuple[int, str]] = None
start = 0
for index, (number, text) in enumerate(paras[:8]):
plain = strip_markers(text)
no_end_punct = not any(ch in HEADER_END_PUNCT for ch in plain)
not_heading = not TOP_HEADING_RE.match(text)
if (no_end_punct and not_heading and plain.rstrip().endswith((":", ":"))
and nonspace_len(plain) <= SALUTATION_MAX_CHARS):
self.salutation = (number, text)
start = index
break
if not (nonspace_len(plain) <= 60 and no_end_punct and not_heading):
start = index
break
self.header.append((number, text))
start = index + 1
self.body: List[Tuple[int, str]] = paras[start:]
def char_count(self) -> int:
return sum(nonspace_len(strip_markers(text)) for _n, text in self.body)
def marker_counts(self) -> Dict[str, int]:
counts = {"待补": 0, "待核实": 0}
for _n, text in self.paras:
for match in MARKER_RE.finditer(text):
counts[match.group(1)] += 1
return counts
def sections(self, rate: float) -> List[Dict[str, object]]:
"""按一级标题(一、二、……)分段统计字数;第一个一级标题之前记为“开场”。"""
result: List[Dict[str, object]] = []
current = {"heading": "开场", "line": self.body[0][0] if self.body else None, "chars": 0}
for number, text in self.body:
if TOP_HEADING_RE.match(text):
if current["chars"] or current["heading"] != "开场":
result.append(current)
current = {"heading": clip(text, 20), "line": number, "chars": 0}
current["chars"] = int(current["chars"]) + nonspace_len(strip_markers(text))
if current["chars"] or current["heading"] != "开场":
result.append(current)
for item in result:
item["minutes"] = round(int(item["chars"]) / rate, 1)
return result
# ---------------------------------------------------------------------------
# 检查
# ---------------------------------------------------------------------------
def make_issue(issue_id: str, level: str, category: str, line: Optional[int], excerpt: str,
message: str, suggestion: str, basis: str) -> Dict[str, object]:
return {
"id": issue_id,
"level": level,
"category": category,
"line": line,
"excerpt": excerpt,
"message": message,
"suggestion": suggestion,
"basis": basis,
}
def count_numbers(text: str) -> int:
count = 0
for match in NUMBER_RE.finditer(text):
if DATE_TAIL_RE.match(text[match.end():]):
continue
count += 1
return count
def find_quotes(number: int, text: str) -> Tuple[List[Dict[str, object]], List[Dict[str, object]]]:
"""返回(引语清单,问题)。"""
quotes: List[Dict[str, object]] = []
issues: List[Dict[str, object]] = []
for match in QUOTE_RE.finditer(text):
before = text[:match.start()]
if not QUOTE_VERB_RE.search(before):
continue
# 本句的起点:上一个句末标点之后
sentence_start = max(before.rfind(ch) for ch in SENTENCE_END) + 1
lead = before[sentence_start:].lstrip("”’))」』 ,,、")
after = text[match.end():match.end() + 40]
has_source = bool(SOURCE_RE.search(lead))
marked = "【待核实" in lead or "【待核实" in after.split("。")[0]
content = match.group(1)
quotes.append({
"line": number,
"lead": clip(lead, 30),
"text": clip(content, 40),
"has_source": has_source,
"marked": marked,
})
vague = bool(VAGUE_SOURCE_P1_RE.search(lead) or VAGUE_SOURCE_P2_RE.search(lead))
if not has_source and not marked and not vague: # 来源不明的引子另由 QUOTE-002、003 报告
issues.append(make_issue(
"QUOTE-001", "P2", "引语", number, clip(lead + "“" + content + "”", 40),
"引语未注明出处,也没有标【待核实】",
"核对是否为用户提供的原文:是的注明出处,不是的删去或改为转述并标【待核实:请对照原文】",
"references/_shared/guardrails.md 第二条第1、5款",
))
for match in VAGUE_SOURCE_P1_RE.finditer(text):
issues.append(make_issue(
"QUOTE-002", "P1", "引语", number, clip(text[match.start():], 30),
f"“{match.group(1)}”引出的话来源不明,讲话稿中的引文只能来自用户提供的原文",
"有用户提供的原文和出处的,改为注明出处的引用;没有的删去引子,用自己的话直接说",
"references/_shared/guardrails.md 第二条第5款",
))
for match in VAGUE_SOURCE_P2_RE.finditer(text):
issues.append(make_issue(
"QUOTE-003", "P2", "引语", number, clip(text[match.start():], 30),
f"“{match.group(1)}”引出的话请核对来源:是否为用户提供的原文,或同场讲话中确实说过",
"核对属实的保留;核对不了的删去引子,用自己的话直接说",
"references/_shared/guardrails.md 第二条第5款",
))
return quotes, issues
def check(speech: Speech, minutes: Optional[float], rate: float, style: str,
max_sentence: int, max_numbers: int, max_para: int) -> Dict[str, object]:
issues: List[Dict[str, object]] = []
quotes: List[Dict[str, object]] = []
chars = speech.char_count()
est = round(chars / rate, 1)
first_line = speech.body[0][0] if speech.body else None
basis_time = "工具约定:按每分钟约 200 字换算,误差±10%(references/oral-style.md 第五部分)"
target_range: Optional[List[int]] = None
within: Optional[bool] = None
if minutes is not None:
target = minutes * rate
low, high = int(round(target * (1 - TOLERANCE))), int(round(target * (1 + TOLERANCE)))
target_range = [low, high]
within = low <= chars <= high
if not within:
direction = "超出" if chars > high else "不足"
if chars > high:
tip = "按铺垫、重复表态、数字罗列、次要事例的顺序删减"
else:
tip = "只展开要点清单中的事实;素材不够的按实际篇幅交付,在起草说明写明差距"
issues.append(make_issue(
"TIME-001", "P1", "时长", first_line, f"正文约 {chars} 字",
f"预计 {est} 分钟,目标 {minutes:g} 分钟({low}~{high} 字),篇幅{direction}",
tip, basis_time,
))
if est > HOUR_LIMIT_MINUTES:
issues.append(make_issue(
"TIME-002", "P1", "时长", first_line, f"正文约 {chars} 字",
f"预计 {est} 分钟,超过 1 小时",
"主要负责同志讲话、会议主报告不超过 1 小时,请压缩;用户坚持的,在起草说明提示",
"《整治形式主义为基层减负若干规定》第6条",
))
if speech.salutation is None:
issues.append(make_issue(
"ORAL-004", "P2", "称谓", first_line, clip(speech.body[0][1]) if speech.body else "",
"开头没有找到单独成段的称谓(如“同志们:”)",
"在开场前加称谓并单独成段;致辞按名单的礼宾顺序称呼",
"GB/T 15834-2011 4.7.3.4;references/types/ceremony.md 第五部分",
))
later_salutations = 0
for index, (number, text) in enumerate(speech.body):
plain = strip_markers(text)
if speech.salutation is not None and index > 0:
later_salutations += len(SALUTATION_WORDS_RE.findall(plain))
para_len = nonspace_len(plain)
if para_len > max_para:
issues.append(make_issue(
"ORAL-003", "P2", "口语化", number, clip(text),
f"本段 {para_len} 字,超过 {max_para} 字",
"拆成两三段,每段讲一层意思", "工具约定(references/oral-style.md 第二、三部分)",
))
numbers = count_numbers(LIST_NUMBER_RE.sub("", plain, count=1))
if numbers > max_numbers:
issues.append(make_issue(
"ORAL-002", "P2", "口语化", number, clip(text),
f"本段连用 {numbers} 个数字,超过 {max_numbers} 个,听众记不住",
"留最能说明问题的一两个,其余删去或写进书面材料;数据不取整",
"工具约定(references/oral-style.md 第二部分第3条)",
))
is_salutation = speech.salutation is not None and index == 0
for match in ([] if is_salutation else SENTENCE_RE.finditer(plain)): # 称谓行不按整句检查
sentence = match.group(0).rstrip(SENTENCE_END)
length = nonspace_len(sentence)
if length > max_sentence:
issues.append(make_issue(
"ORAL-001", "P2", "口语化", number, clip(sentence),
f"整句 {length} 字,超过 {max_sentence} 字,不便口头表达",
"拆成两三句,一句一个意思;长定语拆开,动词提前",
"工具约定(references/oral-style.md 第二部分第1、2条)",
))
found_quotes, quote_issues = find_quotes(number, text)
quotes.extend(found_quotes)
issues.extend(quote_issues)
if (style == "讲" and speech.salutation is not None and chars > SALUTATION_HINT_CHARS
and later_salutations == 0):
issues.append(make_issue(
"ORAL-005", "P2", "称谓", speech.salutation[0], clip(speech.salutation[1]),
"全篇只在开头称呼一次",
"讲的稿在段落转折、提出重要要求时用称谓唤起注意,全篇三到五次",
"工具约定(references/oral-style.md 第二部分第4条)",
))
summary = {"P0": 0, "P1": 0, "P2": 0}
for issue in issues:
summary[str(issue["level"])] += 1
return {
"char_count": chars,
"rate": int(rate) if float(rate).is_integer() else rate,
"style": style,
"est_minutes": est,
"target_minutes": minutes,
"target_range": target_range,
"within_target": within,
"markers": speech.marker_counts(),
"salutation": speech.salutation[1] if speech.salutation else None,
"header": [text for _n, text in speech.header],
"sections": speech.sections(rate),
"quotes": quotes,
"summary": summary,
"issues": issues,
"thresholds": {"max_sentence": max_sentence, "max_numbers": max_numbers, "max_para": max_para},
}
# ---------------------------------------------------------------------------
# 输出
# ---------------------------------------------------------------------------
def render_text(result: Dict[str, object]) -> str:
lines: List[str] = []
style_name = "讲的稿" if result["style"] == "讲" else "念的稿"
lines.append(f"讲话稿检查:{result['file']}({style_name},语速每分钟约 {result['rate']:g} 字)")
head = f"正文约 {result['char_count']} 字,预计 {result['est_minutes']} 分钟"
if result["target_minutes"] is not None:
low, high = result["target_range"] # type: ignore[misc]
if result["within_target"]:
verdict = "在范围内"
else:
verdict = "篇幅超出" if result["char_count"] > high else "篇幅不足"
head += f";目标 {result['target_minutes']:g} 分钟({low}~{high} 字),{verdict}"
lines.append(head + "。")
markers = result["markers"] # type: ignore[assignment]
if markers["待补"] or markers["待核实"]: # type: ignore[index]
lines.append(f"待补 {markers['待补']} 处、待核实 {markers['待核实']} 处,补齐后字数会变化。") # type: ignore[index]
if result.get("truncated_line"):
lines.append(f"第 {result['truncated_line']} 行为【起草说明】,其后内容未检查。")
sections = result["sections"] # type: ignore[assignment]
if sections:
parts = [f"{s['heading']} {s['chars']} 字({s['minutes']} 分钟)" for s in sections] # type: ignore[index]
lines.append("分段:" + ";".join(parts))
summary = result["summary"] # type: ignore[assignment]
lines.append(f"问题:P1 {summary['P1']} 条,P2 {summary['P2']} 条") # type: ignore[index]
for issue in result["issues"]: # type: ignore[union-attr]
where = f"第{issue['line']}行" if issue["line"] else ""
lines.append(f" [{issue['level']}] {issue['id']} {where}:{issue['message']}。原文:{issue['excerpt']}")
lines.append(f" 建议:{issue['suggestion']}(依据:{issue['basis']})")
quotes = result["quotes"] # type: ignore[assignment]
if quotes:
lines.append(f"引语 {len(quotes)} 处,请逐条对照用户提供的原文和出处,整理进引语核对清单:")
for item in quotes: # type: ignore[union-attr]
if item["marked"]:
state = "已标待核实"
elif item["has_source"]:
state = "已写出处"
else:
state = "未写出处"
lines.append(f" 第{item['line']}行:{item['lead']}“{item['text']}”({state})")
return "\n".join(lines)
def build_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(
prog="speech_check.py",
description="讲话稿时长与口语化检查:估算时长,找出长句、数字过密的段落和来源不明的引文。",
epilog="退出码:0 正常;--strict 且存在 P1 时为 1;输入错误为 2。docx 输入时“行号”为段落序号。",
add_help=False,
formatter_class=_ChineseHelpFormatter,
)
parser._positionals.title = "位置参数"
parser._optionals.title = "选项"
parser.add_argument("-h", "--help", action="help", help="显示本帮助信息并退出")
parser.add_argument("file", help="讲话稿(.txt、.md、.docx),遇到【起草说明】行即停止读取")
parser.add_argument("--minutes", type=float, default=None, help="目标时长(分钟),给出时检查字数是否在目标的±10%%以内")
parser.add_argument("--rate", type=float, default=DEFAULT_RATE, help="语速,每分钟字数,默认 200")
parser.add_argument("--style", choices=("讲", "念"), default="讲",
help="稿子用法:讲(讲的稿,默认)或念(念的稿,阈值放宽)")
parser.add_argument("--max-sentence", type=int, default=None, help="整句字数上限,默认讲 40、念 60")
parser.add_argument("--max-numbers", type=int, default=None, help="每段数字个数上限,默认讲 3、念 5")
parser.add_argument("--max-para", type=int, default=None, help="每段字数上限,默认讲 250、念 400")
parser.add_argument("--json", action="store_true", help="输出 JSON")
parser.add_argument("--strict", action="store_true", help="存在 P1 时返回 1")
parser.add_argument("--version", action="version", version=f"%(prog)s {VERSION}", help="显示版本号")
return parser
def run(args: argparse.Namespace) -> Dict[str, object]:
if args.minutes is not None and args.minutes <= 0:
raise InputError("--minutes 必须大于 0")
if args.rate <= 0:
raise InputError("--rate 必须大于 0")
defaults = STYLE_DEFAULTS[args.style]
max_sentence = args.max_sentence if args.max_sentence is not None else defaults["max_sentence"]
max_numbers = args.max_numbers if args.max_numbers is not None else defaults["max_numbers"]
max_para = args.max_para if args.max_para is not None else defaults["max_para"]
for name, value in (("--max-sentence", max_sentence), ("--max-numbers", max_numbers),
("--max-para", max_para)):
if value <= 0:
raise InputError(f"{name} 必须大于 0")
path = Path(args.file)
speech = Speech(read_lines(path))
if not speech.body:
raise InputError("没有读到正文内容。")
result = check(speech, args.minutes, args.rate, args.style, max_sentence, max_numbers, max_para)
ordered: Dict[str, object] = {"file": path.name, "truncated_line": speech.truncated_line}
ordered.update(result)
return ordered
def main(argv: Optional[Sequence[str]] = None) -> int:
setup_console()
parser = build_parser()
try:
args = parser.parse_args(argv)
except SystemExit as exc: # argparse 参数错误返回 2,--help 返回 0
return int(exc.code or 0)
try:
result = run(args)
except InputError as exc:
print(f"输入错误:{exc}", file=sys.stderr)
return 2
if args.json:
print(json.dumps(result, ensure_ascii=False, indent=2))
else:
print(render_text(result))
summary = result["summary"] # type: ignore[assignment]
if args.strict and summary["P1"]: # type: ignore[index]
return 1
return 0
if __name__ == "__main__":
sys.exit(main())
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |