wordcloud标准教程

    xiaoxiao2022-06-30  155

    # -*- coding: utf-8 -*- """ """ from os import path from scipy.misc import imread from wordcloud import WordCloud, STOPWORDS import matplotlib.pyplot as plt def wordCount(filename): ''' 简单计算词频的函数 :param filename: 文件名 :return: 词频 ''' wordCount = {} file = open(filename,'r') while True: line = file.readline() if line: wordlist = line.split(',[') if len(wordlist) == 1: continue wordlist = wordlist[1].split('],')[0].split(',') for word in wordlist: word = word.replace(' ', '').replace("'", '') if word in wordCount.keys(): wordCount[word] = wordCount[word]+1 else: wordCount[word] = 1 else: break return [(k, wordCount[k]) for k in wordCount.keys()] def generateCloud(filename,imagename,cloudname,fontname): ''' 生成标签云的函数 ''' coloring = imread(imagename) # 读取背景图片 wc = WordCloud(background_color="white", # 背景颜色max_words=2000,# 词云显示的最大词数 mask=coloring, # 设置背景图片 stopwords=STOPWORDS, # 停止词 font_path=fontname, # 兼容中文字体 max_font_size=150) # 字体最大值 #计算好词频后使用generate_from_frequencies函数生成词云 #txtFreq例子为[('词a', 100),('词b', 90),('词c', 80)] txtFreq = wordCount(filename) wc.generate_from_frequencies(txtFreq) # 生成图片 plt.imshow(wc) plt.axis("off") # 绘制词云 plt.figure() # 保存词云 wc.to_file(cloudname) if __name__ == '__main__': d = path.dirname(__file__) # 获取当前文件路径 fontname = path.join(d, 'msyh.ttf') # 中文字体路径 filename = path.join(d, '广州.txt') # txt文件路径 imagename = path.join(d, "circle.jpg") # 背景图片路径 cloudname = path.join(d, "cloud.png") # 标签云路径 generateCloud(filename, imagename, cloudname, fontname)

    最新回复(0)