[Notes] Stuttering word segmentation to draw word cloud map

一、结巴分词的三种模式

（1）精确模式：Cut the words that are most likely to make up the words,No redundant words.

（2）全模式：Cut all possible words into words,There are redundant words.

（3）搜索引擎模式：在精确模式的基础上,对长词再次切分,适合用于搜索引擎分词.

二、正则提取数据

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

# 导数据
douyin = pd.read_csv('data/douyin.csv')

# 正则提取,并达到MySQL中group_concat的效果
temp = douyin['signature'].str.extractall(r'[^一-龥]*([一-龥]+)[^一-龥]*').copy()
temp = temp.reset_index().groupby('level_0').agg({
    0:list})[0].apply(lambda x: ','.join(x))

# Tokenize the extracted string,并剔除停用词
result0 = ','.join([','.join(jieba.lcut(statement)) for statement in temp]).split(',')
result1 = [x for x in result0 if x not in stop_words]

# Create word frequenciesSeries对象
important_words = pd.Series(result1).value_counts()[1:200]
important_words

三、绘制

# Import reference images required for drawing
bgimg = np.array(Image.open('data/bgimg.png'))
# Use the color of the reference image as the color of the word cloud
genclr = wordcloud.ImageColorGenerator(bgimg)

# Create a Ciyun map object
wc = wordcloud.WordCloud(
                        # 指定字体路径
                        font_path='data/FZZJ-LongYTJW.TTF',
                        # 指定背景颜色
                        background_color='white',
                        # Specifies the maximum number of words
                        max_words=200,
                        # Specify the maximum and minimum font size
                        max_font_size=300,
                        min_font_size=5,
                        # 指定随机种子
                        random_state=4,
                        # Specifies which graph the word cloud graph outline refers to
                        mask=bgimg,
                        # Specifies the word cloud color
                        color_func=genclr)
# 渲染文字
wc.generate_from_frequencies(important_words)
# 使用plt展示词云图
plt.figure(figsize=(24,24))
plt.imshow(wc)
plt.axis('off')