単語長分布ソースコード

word_distribution.py

# encoding: 'utf-8'

import random
import numpy as np
from datetime import datetime
from math import factorial, exp

SYL_MAX = 50
SAMPLING = 20
M = 0
N = 6350

def volume(i):
    if i <= 2:
        return 1+i
    return 6*i

def buffer_effect(i, word):
    v = volume(i)
    if v == 1:
        return 1

    result = 1
    for k in range(1, v):
        result *= (space(i) - word*v - k) / (space(i) - word*(v-1) - k)
    return result

def quality_filter(i, word, total):
    S = 0.5
    return (1 - S)


def space(i):
    # return M**i
    # return factorial(M) / (factorial(M-i) * factorial(i))
    # return int(M / i) ** i
    # return int(250 * (4*27)**(i-1) / i**i)
    return 250 * 60 ** (i-1)
    # return 92 * 36 ** (i-1)
    # return 870**i


def run(total):
    words = [0] * SYL_MAX
    for i in range(N):
        for j in range(SYL_MAX):
            r = random.randint(1, SPACE[j])
            word = words[j]
            boundary = (SPACE[j] - volume(j+1) * word) * buffer_effect(j+1, word) * quality_filter(j+1, word, total)
            if r < boundary:
                words[j] += 1
                break
            else:
                continue
    return words


data = list()
SPACE = [space(i+1) for i in range(SYL_MAX)]


if __name__ == '__main__':
    # print(SPACE)
    for i in range(SAMPLING):
        print("SAMPLING: {}".format(i+1))
        data.append(run(i+1))

    data = np.array(data)
    average = np.average(data, axis=0)

    tdatetime = datetime.now()
    tstr = tdatetime.strftime('%Y%m%d%H%M%S')
    np.savetxt("output/{}_rawdata.csv".format(tstr), data.T, delimiter=',')
    np.savetxt("output/{}_average.csv".format(tstr), average.T, delimiter=',')

    print("DONE")