关于「东西方原典考辨:《墨子·非命上》三表法与加菲尔德的引文索引」,《生成式搜索引用权重》给出的定量结论是什么?

结论:第七章考辨《墨子·非命上》三表法与加菲尔德 1955 年的引文索引:累积优势使质量与被引次数的秩相关从 0.8688 降到 0.6773。

数据口径:书中参数均为示意;全部计算由书中程序生成。 第五章讲可核查性,并用模拟说明从外部推断特征权重的脆弱:漏掉长度时统计数字的权重被高估 0.3252。

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: 0.6773, 0.8688

pages 2000, citations 20000
world top 10% max zero rho(q,c)
Q quality only0.2116 31 34 0.8688
C cumulative0.5169 283 441 0.6773
same pages, two reruns of world C: rho(c1,c2)0.4848
查看完整验算程序
# Chapter 7. p72: citation counts measure use, and use feeds on itself.
# Garfield (1955) proposed indexing who cites whom so that a reader can
# follow the use of an idea. We simulate 2000 answer pages with an
# integer quality q from 1 to 9 and hand out 20000 citations one at a
# time. World Q: chance proportional to q. World C: proportional to
# q (c + 1), a cumulative advantage in the spirit of Price (1976).
# We report concentration and the rank correlation between quality
# and final citations. Parameters are illustrative; seed is fixed.
import math
import random

rng = random.Random(20261113)
PAGES, CITES = 2000, 20000
QUAL = [rng.randint(1, 9) for _ in range(PAGES)]


class Fenwick:
    def __init__(self, n):
        self.n, self.t = n, [0] * (n + 1)

    def add(self, i, v):
        i += 1
        while i <= self.n:
            self.t[i] += v
            i += i & -i

    def find(self, x):
        pos, step = 0, 1 << self.n.bit_length()
        while step:
            nxt = pos + step
            if nxt <= self.n and self.t[nxt] <= x:
                pos, x = nxt, x - self.t[nxt]
            step >>= 1
        return pos


def run(feedback, seed):
    r = random.Random(seed)
    c = [0] * PAGES
    fw = Fenwick(PAGES)
    total = 0
    for i, q in enumerate(QUAL):
        fw.add(i, q)
        total += q
    for _ in range(CITES):
        i = fw.find(r.randrange(total))
        c[i] += 1
        if feedback:
            fw.add(i, QUAL[i])
            total += QUAL[i]
    return c


def avg_ranks(xs):
    order = sorted(range(len(xs)), key=lambda i: xs[i])
    rk = [0.0] * len(xs)
    j = 0
    while j < len(order):
        k = j
        while k + 1 < len(order) and xs[order[k + 1]] == xs[order[j]]:
            k += 1
        for m in range(j, k + 1):
            rk[order[m]] = (j + k) / 2 + 1
        j = k + 1
    return rk


def spearman(a, b):
    ra, rb = avg_ranks(a), avg_ranks(b)
    ma, mb = math.fsum(ra) / len(ra), math.fsum(rb) / len(rb)
    cov = math.fsum((x - ma) * (y - mb) for x, y in zip(ra, rb))
    va = math.fsum((x - ma) ** 2 for x in ra)
    vb = math.fsum((y - mb) ** 2 for y in rb)
    return cov / math.sqrt(va * vb)


print("pages %d, citations %d" % (PAGES, CITES))
print("%-22s %9s %9s %9s %8s" % ("world", "top 10%", "max", "zero",
                                 "rho(q,c)"))
for name, fb in (("Q quality only", False), ("C cumulative", True)):
    c = run(fb, 7)
    top = sorted(c, reverse=True)[:PAGES // 10]
    share = math.fsum(top) / CITES
    zero = len([x for x in c if x == 0])
    print("%-22s %9.4f %9d %9d %8.4f" % (name, share, max(c), zero,
                                         spearman(QUAL, c)))
c1, c2 = run(True, 7), run(True, 8)
print("same pages, two reruns of world C: rho(c1,c2) = %.4f" %
      spearman(c1, c2))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《生成式搜索引用权重几何学、问答引擎检索重排特征与高被引双语直答页段落拓扑母典》第 7 章「东西方原典考辨:《墨子·非命上》三表法与加菲尔德的引文索引」(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-260 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026260,
  author    = {Fan, Yafu},
  title     = {The Geometry of Citation Weight in Generative Search: Retrieval and Reranking Features and the Paragraph Topology of Highly Cited Bilingual Answer Pages},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论