关于「歧义熵:每条并置边上的读法分布与读者负担」,《字源、意合与中英双轨》给出的定量结论是什么?

结论:把读法的分布算进来,意合句的歧义就有了数值。《论语》开篇一章合计 8.273 比特,温庭筠一联 7.806 比特,平均每条边分别是 1.182 与 1.115 比特。两位读者全句读成一样的概率不到百分之一,平均在三条多的边上分歧。加一个虚词最值钱的位置,是熵最高的那条边;加到第三个,留白只剩 0.571。

数据口径:以今律古按示意参数每千字有 1.307 次期望误读,错句只有三句时抽检须读全稿的 0.632。 书中参数为示意,全部计算由书中程序生成(23 段程序)。

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 标准库与结论吻合的打印值: 1.115, 1.182, 7.806, 8.273

max per edge: log2(5)2.32 bits
Lunyu 1.1
learn-practise (er)1.219 mode coord marked
practise->glad1.578 mode cond
friends->joy1.369 mode cause
unknown-unvexed (er)0.922 mode contrast marked
unvexed->noble1.361 mode cause
question1-20.748 mode coord
question2-31.076 mode coord
total H 8.273 bits of max 16.25; share 0.509; 2**H309.4
Wen couplet
crow-inn1.539 mode time
inn-moon0.971 mode coord
steps-bridge0.881 mode coord
bridge-frost0.934 mode coord
crow||steps1.353 mode coord
inn||bridge0.722 mode coord
moon||frost1.406 mode coord
total H 7.806 bits of max 16.25; share 0.480; 2**H223.9
per-edge mean: Lunyu 1.182, Wen 1.115
edges above 1.40 bits: Lunyu 1
edges above 1.40 bits: Wen 2
查看完整验算程序
# Chapter 4. p41: ambiguity entropy of each relation edge
# Five relation types: cause, contrast, coordination (incl. co-presence),
# time, condition. Each edge carries the author's illustrative reading
# distribution (to be replaced by reader-panel counts). H(edge) in bits;
# H(sentence) adds edges as if independent; 2 ** H is the number of
# equally likely readings with the same uncertainty.
import math

TYPES = ("cause", "contrast", "coord", "time", "cond")
LUNYU = [
    ("learn-practise (er)", (0.05, 0.00, 0.55, 0.40, 0.00), True),
    ("practise->glad", (0.35, 0.00, 0.05, 0.10, 0.50), False),
    ("friends->joy", (0.45, 0.00, 0.00, 0.10, 0.45), False),
    ("unknown-unvexed (er)", (0.00, 0.80, 0.10, 0.00, 0.10), True),
    ("unvexed->noble", (0.50, 0.00, 0.10, 0.00, 0.40), False),
    ("question1-2", (0.00, 0.05, 0.85, 0.10, 0.00), False),
    ("question2-3", (0.00, 0.25, 0.70, 0.05, 0.00), False),
]
WEN = [
    ("crow-inn", (0.30, 0.00, 0.25, 0.45, 0.00), False),
    ("inn-moon", (0.00, 0.00, 0.60, 0.40, 0.00), False),
    ("steps-bridge", (0.00, 0.00, 0.70, 0.30, 0.00), False),
    ("bridge-frost", (0.00, 0.00, 0.65, 0.35, 0.00), False),
    ("crow||steps", (0.15, 0.00, 0.60, 0.25, 0.00), False),
    ("inn||bridge", (0.00, 0.20, 0.80, 0.00, 0.00), False),
    ("moon||frost", (0.00, 0.15, 0.55, 0.30, 0.00), False),
]


def entropy(p):
    return -math.fsum(x * math.log2(x) for x in p if x > 0)


def table(name, edges):
    print(name)
    hs = []
    for label, p, marked in edges:
        h = entropy(p)
        hs.append(h)
        print("  {:<22}{:>6.3f}  mode {:<6}{}".format(
            label, h, TYPES[p.index(max(p))], "  marked" if marked else ""))
    tot = math.fsum(hs)
    hmax = len(edges) * math.log2(len(TYPES))
    print("  total H {:.3f} bits of max {:.2f}; share {:.3f}; "
          "2**H = {:.1f}".format(tot, hmax, tot / hmax, 2 ** tot))
    return tot


print("max per edge: log2(5) = {:.2f} bits".format(math.log2(5)))
a = table("Lunyu 1.1", LUNYU)
b = table("Wen couplet", WEN)
print("per-edge mean: Lunyu {:.3f}, Wen {:.3f}".format(a / 7, b / 7))
for name, edges in (("Lunyu", LUNYU), ("Wen", WEN)):
    k = 0
    for _, p, _ in edges:
        if entropy(p) > 1.40:
            k += 1
    print("edges above 1.40 bits: {} {}".format(name, k))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《古汉语字源本义训诂、意合留白拓扑与中西意境张力转化母典》第 4 章「歧义熵:每条并置边上的读法分布与读者负担」(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-382 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026382,
  author    = {Fan, Yafu},
  title     = {Classical Chinese Etymological Philology, Paratactic Imagery Topology and East-West Aesthetic Asymmetry in High-Authority Writing},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论