关于「形合树与意合网:留白比例与并置边的图结构」,《字源、意合与中英双轨》给出的定量结论是什么?
结论:第三至六章把句子画成图,定义留白比例与每条并置边的歧义熵(《论语》开篇一章合计 8.273 比特),在英文可追溯率不低于 0.90 的约束下穷举最优方案:温庭筠一联最多保留 0.314 的留白,《论语》开篇一章保留 0.371;参数扰动下最优方案守住红线的概率只有 0.598。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。
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)。
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
| PDTB 2.0 relation tokens | 40,600 |
| Explicit | 18,459 0.4547 |
| Implicit | 16,053 0.3954 |
| AltLex | 624 0.0154 |
| EntRel | 5,210 0.1283 |
| NoRel | 254 0.0063 |
| not explicit | 22,141 = 0.5453 (+/- 0.0048 at 95%) |
| implicit among explicit+implicit | 0.4651 |
| CTB rough count 0.82 vs PDTB 0.5453: odds ratio 3.80 | |
| text edges marked gap ratio | |
| Lunyu 1.1 | 7 2 0.714 |
| Wen couplet | 7 0 1.000 |
| Tianjingsha | 10 0 1.000 |
| Wen, hypotactic English | 5 4 0.200 |
查看完整验算程序
# Chapter 3. p31: how much of discourse is left unmarked?
# Part A, published counts: Penn Discourse TreeBank 2.0 relation tokens
# from the LDC2008T05 readme (Prasad et al. 2008); the Chinese figure
# 0.82 is Zhou and Xue's (2012) rough count over 20 Chinese TreeBank
# files of modern newswire, compared there with 54.5 per cent for PDTB.
# Part B, the book's own "gap ratio" (unmarked relation edges / all
# relation edges) for three public-domain texts, hand-annotated.
import math
PDTB = [("Explicit", 18459), ("Implicit", 16053), ("AltLex", 624),
("EntRel", 5210), ("NoRel", 254)]
tot = 0
for _, k in PDTB:
tot += k
print("PDTB 2.0 relation tokens: {:,}".format(tot))
for name, k in PDTB:
print(" {:<9}{:>7,} {:.4f}".format(name, k, k / tot))
expl = PDTB[0][1]
non = tot - expl
p_en = non / tot
se = math.sqrt(p_en * (1 - p_en) / tot)
print("not explicit: {:,} = {:.4f} (+/- {:.4f} at 95%)".format(
non, p_en, 1.96 * se))
imp_only = PDTB[1][1] / (PDTB[0][1] + PDTB[1][1])
print("implicit among explicit+implicit: {:.4f}".format(imp_only))
p_zh = 0.82
odds = (p_zh / (1 - p_zh)) / (p_en / (1 - p_en))
print("CTB rough count 0.82 vs PDTB {:.4f}: odds ratio {:.2f}".format(
p_en, odds))
print()
GRAPHS = [("Lunyu 1.1", 7, 2), ("Wen couplet", 7, 0),
("Tianjingsha", 10, 0), ("Wen, hypotactic English", 5, 4)]
print("text edges marked gap ratio")
for name, e, m in GRAPHS:
print("{:<25} {:>5} {:>7} {:>10.3f}".format(name, e, m, (e - m) / e))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 3 章「形合树与意合网:留白比例与并置边的图结构」(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}
}