关于「口语化压缩:双人播客的广度与重复、换手时长与香农命名」,《专著改编的信息保留代数》给出的定量结论是什么?

结论:第二章显示三十分钟的播客只讲约 10 个论点时保留 0.4820,把 20 个论点各讲一遍只保留 0.3117,单人连续讲述的最优时长约 41.5 秒。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的经营或投资建议。 第六章在 40 个制作工时下求出播客、投影片与题库的最优组合,价值 2.0685,不做题库只有 1.4164;事实核对应放在上游;每周 2 次提醒使期望学习次数最高,为 6.2792,过量提醒有害,对脆弱学习者阈值更低;只改编有权改编的书,合成内容必须标识。

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: 0.3117, 0.4820, 10, 20

ideasshare of bookretained weightexposures (top, last)
20.05000.26042018
40.10000.3801108
60.15000.445775
80.20000.475263
100.25000.482052
120.30000.476351
140.35000.462451
160.40000.440841
180.45000.402231
200.50000.311711
best breadth 10 ideas, retained weight 0.4820
stating all 40 once would need 60.0 minutes
查看完整验算程序
# Chapter 2. p21: breadth against repetition in a 30-minute two-host show.
# The monograph holds 40 ideas with importance w_i proportional to
# 1 / i^0.8. A listener cannot rewind, so an idea stated once is kept with
# probability q = 0.40; each restatement (an example, a question from the
# second host, a recap) is another independent chance, so k exposures
# keep it with 1 - (1 - q)^k. Stating an idea costs 1.5 minutes, each
# restatement 0.75. For every breadth n (top n ideas) the spare minutes
# go greedily to the restatement with the largest weighted gain; the
# objective is separable and concave in k, so greedy is exact here.
# Times, q and the importance curve illustrative.
import math

N_IDEAS, MINUTES, FIRST, AGAIN, Q = 40, 30.0, 1.5, 0.75, 0.40
W = [1.0 / (i + 1) ** 0.8 for i in range(N_IDEAS)]
TOTAL = math.fsum(W)


def keep(k):
    return 1.0 - (1.0 - Q) ** k


def best_for(n):
    spare = MINUTES - n * FIRST
    if spare < 0:
        return None
    k = [1] * n
    for _ in range(int(spare / AGAIN + 1e-9)):
        j = max(range(n), key=lambda i: W[i] * (keep(k[i] + 1)
                                                - keep(k[i])))
        k[j] += 1
    kept = math.fsum(W[i] * keep(k[i]) for i in range(n))
    return kept / TOTAL, k


print("ideas  share of book  retained weight  exposures (top, last)")
best = (0.0, 0)
for n in (2, 4, 6, 8, 10, 12, 14, 16, 18, 20):
    r = best_for(n)
    if r is None:
        continue
    frac, k = r
    best = max(best, (frac, n))
    print("{:5d}  {:13.4f}  {:15.4f}  {:9d} {:9d}".format(
        n, n / N_IDEAS, frac, k[0], k[-1]))
print("best breadth {} ideas, retained weight {:.4f}".format(
    best[1], best[0]))
print("stating all 40 once would need {:.1f} minutes".format(
    N_IDEAS * FIRST))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《一书多吃内容裂变代数:从硬核专著自动编译双人播客、导学投影片与通关题库》第 2 章「口语化压缩:双人播客的广度与重复、换手时长与香农命名」(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-151 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026151,
  author    = {Fan, Yafu},
  title     = {Content Fission Algebra: Compiling Hard-Core Monographs into Two-Host Podcasts, Study Slides and Quiz Banks},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论