关于「刘勰、戴震与克努特:可核验写作的古老规矩」,《计算著作学与自校验专著》给出的定量结论是什么?

结论:真确的释义在每一处吻合的机会是 0.95,貌似有理的错误释义是 0.50。从 0.30 的先验出发,一处吻合只把信念提到 0.4488;四处到 0.8481;八处到 0.9864;要到 0.99 需要 9 处。错误的释义通过八处检验的机会是 0.00391。第六章的双语门禁是同样的检验,每个数字都有两「处」:中文与英文正文必须印出同一个数字集合。

数据口径:书中的读者、书、错误率与成本均为示意模型;全部计算由书中程序生成;本书不是法律、财务或出版建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。

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。 与结论吻合的打印值: 0.00391, 0.4488, 0.8481, 0.9864, 0.99

passagesP(true | all fit)false gloss survives
10.44880.50000
20.60740.25000
40.84810.06250
80.98640.00391
160.99990.00002
passages needed for 0.99: 9
查看完整验算程序
# Chapter 7. p14: "Without evidence, no belief", as arithmetic. Dai Zhen
# asked that the sense of a word be tested in every passage where it
# occurs. Two rival glosses of a word: the true one fits each passage
# with chance 0.95 (copying errors), a plausible false one with chance
# 0.50. Prior on the true gloss 0.30. Posterior after n passages that all
# fit, and the chance a false gloss survives n passages. The same test
# is what the bilingual gate does: one number, two languages, n chapters.

def posterior(n, prior=0.30, t=0.95, f=0.50):
    a = prior * t ** n
    b = (1 - prior) * f ** n
    return a / (a + b)

if __name__ == "__main__":
    print("   passages   P(true | all fit)   false gloss survives")
    for n in (1, 2, 4, 8, 16):
        print("   {:8d}   {:17.4f}   {:20.5f}".format(
            n, posterior(n), 0.5 ** n))
    n = 0
    while posterior(n) < 0.99:
        n += 1
    print("   passages needed for 0.99: {}".format(n))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《计算著作学、先证伪后立界修辞与自校验专著编译器》第 7 章「刘勰、戴震与克努特:可核验写作的古老规矩」(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-101 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026101,
  author    = {Fan, Yafu},
  title     = {Computational Bibliometrics, Falsification-First Rhetoric and Self-Verifying Monograph Compilers},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论