关于「新品牌命名预检闭环与证伪方案」,《品牌撞名三维预检》给出的定量结论是什么?

结论:第八章给出新品牌命名的七步预检流程、蒸馏小工具「品牌撞名预检器」与预注册检验方案,主要检验约需 1820 对名字,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。

数据口径:在一个示意模型中,声音通道占八成时音码的 AUC 为 0.851,高于字母编辑距离的 0.817。 书中参数为示意;全部计算由书中程序生成;本书是非个人化法律建议,商标近似判断以各地官方检索与专业代理意见为准。

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: 1820

planning AUC sound 0.776shape 0.709confusion share 0.171
powerpairsconfusedsimulated power (400 runs)
0.8018203110.825
0.9024404170.938
查看完整验算程序
# Chapter 8. p82: how many rated name pairs the pre-registered test needs.
# Model: the test asks whether the AUC of sound similarity exceeds the
# AUC of shape similarity in a mixed-channel panel. Planning values
# are the p61 results (0.776 and 0.709) and its confusion share
# (684 of 4000). Standard error of each AUC by Hanley and McNeil
# (1982); the two AUCs are treated as independent, which ignores
# their positive correlation and so overstates the n needed. A
# binormal simulation (fixed seed, 400 runs) checks the power.
import math
import random
from statistics import NormalDist

A1, A2 = 0.776, 0.709
PREV = 684 / 4000
Z = NormalDist()


def se(a, n1, n0):
    q1, q2 = a / (2 - a), 2 * a * a / (1 + a)
    v = math.fsum([a * (1 - a), (n1 - 1) * (q1 - a * a),
                   (n0 - 1) * (q2 - a * a)]) / (n1 * n0)
    return math.sqrt(v)


def need(power, alpha=0.05):
    za, zb = Z.inv_cdf(1 - alpha / 2), Z.inv_cdf(power)
    n = 50
    while True:
        n1 = round(n * PREV)
        n0 = n - n1
        s = math.hypot(se(A1, n1, n0), se(A2, n1, n0))
        if (A1 - A2) / s >= za + zb:
            return n, n1
        n += 10


def auc(pos, neg):
    allv = sorted([(v, 1) for v in pos] + [(v, 0) for v in neg])
    ranks = [i + 1.0 for i, (v, y) in enumerate(allv) if y]
    n1, n0 = len(pos), len(neg)
    return (math.fsum(ranks) - n1 * (n1 + 1) / 2.0) / (n1 * n0)


def sim_power(n, n1, runs, rng):
    m1 = math.sqrt(2) * Z.inv_cdf(A1)
    m2 = math.sqrt(2) * Z.inv_cdf(A2)
    za, hits = Z.inv_cdf(0.975), 0
    for _ in range(runs):
        n0 = n - n1
        a = auc([rng.gauss(m1, 1) for _ in range(n1)],
                [rng.gauss(0, 1) for _ in range(n0)])
        b = auc([rng.gauss(m2, 1) for _ in range(n1)],
                [rng.gauss(0, 1) for _ in range(n0)])
        s = math.hypot(se(a, n1, n0), se(b, n1, n0))
        if (a - b) / s >= za:
            hits += 1
    return hits / runs


if __name__ == "__main__":
    print("planning AUC sound {}  shape {}  confusion share {:.3f}"
          .format(A1, A2, PREV))
    rng = random.Random(20261113)
    print("power  pairs  confused  simulated power (400 runs)")
    for pw in (0.80, 0.90):
        n, n1 = need(pw)
        print("{:.2f} {:7d} {:9d} {:12.3f}".format(
            pw, n, n1, sim_power(n, n1, 400, rng)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《商标音近形近义近三维距离、从反切到音码的品牌撞名预检母典》第 8 章「新品牌命名预检闭环与证伪方案」(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-317 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026317,
  author    = {Fan, Yafu},
  title     = {Phonetic, Graphic and Semantic Distance between Trademarks: Pre-Screening Brand-Name Collisions from Fanqie Spellings to Sound Codes},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论