关于「实战工具与检查器规格:嵌词台账、四件套检查器与预注册检验」,《嵌词阈值与单语等效》给出的定量结论是什么?

结论:第八章给出嵌词台账、四件套检查器规格、蒸馏小工具「多语嵌词净收益与单语构式等效试算器」与预注册检验。书中参数为示意,全部计算由书中程序生成,不对应任何真实客户、对话或成交;马来语、闽南语、粤语、英语、华语一律平等对待。本书不是语言学定论,也不构成针对个人或企业的经营建议。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。

数据口径:嵌词收益只在读者共享双语词库时成立:把单语表达、嵌入与嵌入加注释写成覆盖率的三条成本线,示意参数下 kiasu 在书面渠道的覆盖率阈值为 0.844;读者不发问时,失败会变成沉默的误读;各地区分页面值得嵌的词,合成混合页面后可能全部跌破阈值。 书中参数为示意,全部计算由书中程序生成,不对应任何真实客户、对话或成交;马来语、闽南语、粤语、英语、华语一律平等对待。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库

idwordpp*dd_maxcovtransverdict
L1kiasu0.820.8440.82.620.89.5AMBER PSTAR
D2lepak?0.9671.6-0.87.0RED UNK
C3paiseh0.950.9330.45.101.06.0GREEN
B4sajiao0.100.8580.30.000.84.0AMBER PSTAR,DENS
L5shiok0.90-2.54.020.68.0RED COV,NEG
C6followup 0.970.9353.55.561.021.0RED TRS
D7kiasu0.950.8440.55.100.85.0RED LINE
tally: RED 4, AMBER 2, GREEN 1
查看完整验算程序
# Chapter 8. p81: the insertion ledger, seven synthetic rows, one verdict.
# Per row: coverage p (None when unknown), insertions per 100 chars d,
# bits saved for knowers G (p31), failure cost F = Le + K (p41),
# coverage of the monolingual rewrite, translationese per 1,000 chars,
# and whether the copy trips a red line. Rules (sources in brackets):
#   red   p unknown and d > 1.0 [ch5]; rewrite coverage < 0.8 [ch6];
#         translationese > 18.44 [ch6]; copy red line [ch8]
#   amber p < p* = F / (G + F) [ch4]; G <= 0 [ch3]; d > d_max(p) [ch5]
# Rows are synthetic and describe no real customer or message.
import math

A, B, C = 3.0, 0.15, 6.0
ROWS = [
    ("L1", "kiasu", 0.82, 0.8, 7.22, 38.97, 0.8, 9.5, False),
    ("D2", "lepak", None, 1.6, 1.33, 38.55, 0.8, 7.0, False),
    ("C3", "paiseh", 0.95, 0.4, 2.70, 37.55, 1.0, 6.0, False),
    ("B4", "sajiao", 0.10, 0.3, 6.82, 41.29, 0.8, 4.0, False),
    ("L5", "shiok", 0.90, 2.5, -3.80, 39.14, 0.6, 8.0, False),
    ("C6", "followup", 0.97, 3.5, 2.10, 30.00, 1.0, 21.0, False),
    ("D7", "kiasu", 0.95, 0.5, 7.22, 38.97, 0.8, 5.0, True),
]


def d_max(p):
    u = lambda d: A * math.log1p(d) - B * d * d - C * (1 - p) * d
    if A - C * (1 - p) <= 0:
        return 0.0
    lo, hi = 1e-9, 40.0
    for _ in range(200):
        mid = (lo + hi) / 2
        if u(mid) > 0:
            lo = mid
        else:
            hi = mid
    return lo


def verdict(row):
    _, _, p, d, g, f, cov, tr, line = row
    red, amber = [], []
    if p is None and d > 1.0:
        red.append("UNK")
    if cov < 0.8:
        red.append("COV")
    if tr > 18.44:
        red.append("TRS")
    if line:
        red.append("LINE")
    pstar = f / (g + f) if g > 0 else None
    if g <= 0:
        amber.append("NEG")
    elif p is not None and p < pstar:
        amber.append("PSTAR")
    if p is not None and d > d_max(p):
        amber.append("DENS")
    level = "RED" if red else ("AMBER" if amber else "GREEN")
    return level, pstar, red + amber


if __name__ == "__main__":
    print("id  word      p     p*     d   d_max  cov  trans  verdict")
    tally = {}
    for r in ROWS:
        level, pstar, why = verdict(r)
        tally[level] = tally.get(level, 0) + 1
        ps = "  -  " if pstar is None else "{:.3f}".format(pstar)
        pp = "  ?  " if r[2] is None else "{:.2f} ".format(r[2])
        dm = "  -  " if r[2] is None else "{:5.2f}".format(d_max(r[2]))
        tag = (level + " " + ",".join(why)).strip()
        print("{:<3} {:<8} {} {} {:>4.1f} {} {:>4.1f} {:>5.1f}  {}"
              .format(r[0], r[1], pp, ps, r[3], dm, r[6], r[7], tag))
    print("tally: " + ", ".join("{} {}".format(k, tally[k])
                                for k in ("RED", "AMBER", "GREEN")
                                if k in tally))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《多语参杂语码转换信息论、词汇空缺补偿与单语种等效精准表达母典》第 8 章「实战工具与检查器规格:嵌词台账、四件套检查器与预注册检验」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-381 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026381,
  author    = {Fan, Yafu},
  title     = {Information Theory of Code-Switching, Lexical-Gap Compensation and Monolingual Precision Equivalence in Multilingual Markets},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论