WECHATANALYZER

WeChat (Chinese: 微信) has become the most popular instant messager in China. It provides mobile text and voice messaging communication service, and is available on Android, iPhone, BlackBerry, Windows Phone and Symbian phones. According to wikipedia.org, WeChat has 438 million active users as of august 2014, with 70 million outside of China. People use wechat to interact and socialize, sharing information and comments, as well as communicating to colleagues, therefore, it is quite interesting to explore valuable information from wechat historical messages.

HOW TO USE

Download the project, and unzip into a folder. Set the folder as the current work directory, then import wechatAnalyzer.py. The following is an simple example:

import wechatAnalyzer as wa
test = wa.wechatAnalyzer(file_path="test.xlsx")
data = test.loaddata()
test.clockheat(data)
test.relationship(data)
test.attriplot(data)
test.wordcloudplot_focus(data)
test.generatedict()
test.wordcloudplot_all(data,backimage='guitar.png')

DEPENDENCY

wechatAnalyzer requires the following libraries installed:

• pandas
• numpy
• matplotlib
• wordcloud
• jieba

For wechat messages, we expect a lot of Chinese characters, which can not be displayed correctly when generating wordcloud plots with default settings in 'wordcloud' package. The default font used by 'wordcloud' package is 'DroidSansMono.ttf', a true type font by Google, that is apache licensed. The solution is to do the following modification:

1. copy a font file (here just take 'msyh.ttc' as an example, should use your own lisenced font) to the folder of 'wordcloud' module (e.g. 'C:\Python34\Lib\site-packages\wordcloud')
2. under the same folder, edit 'wordcloud.py' and find the line start with 'FONT_PATH = ', replace 'DroidSansMono.ttf' by 'msyh.ttc' at the end of this line.

EXAMPLES

As an exmpale, 'test.xlsx' file stores a group's wechat messages. iPhone wechat messages can be exported by a tool named 'Tongbu Assistant' (Chinese name: 同步助手). To be noted, exported wechat messages must be saved as a .xlsx file with the exact format.

1 wechatAnalyzer generates a plot to show which clock is the hottest one with most messages happening.

test.clockheat(data)

This command will generate a distribution plot showing how many messages happened on each hour, as follows

2 wechatAnalyzer can provide interesting plot to rank pre-defined index for each person. Pre-defined index include: '花痴', '色', '八卦', '红包', '真相党', '龅牙', '求约'.

test.relationship(data)
test.attriplot(data)

3 wechatAnalyzer provides different types of wordcloud plots for each person in the group.

test.wordcloudplot_focus(data)

Basically, wechatAnalyzer provides two types of wordcloud: focus view wordcloud, and global view wordcloud. For focus view wordcloud, wechatAnalyzer only considers people in the group and takes their IDs as the key words to do the statistics. Therefore, focus view wordcloud plot emphasizes the interaction between people in the group. For global view wordcloud, wechanAnalyzer considers all messages for each person, does tokenization, and then generates wordcloud by using 'wordcloud' module. To do global view wordcloud, use the following command,

test.generatedict()
test.wordcloudplot_all(data,backimage='guitar.png')

To get a better tokenization, wechatAnalyzer will take each people ID as a one 'word' and save into customized dictionary (test.generatedict()). If it failed to generate user dictionary, the default dictionary will be loaded but the tokenization could be bad.

In addition, a file named 'relationship.csv' will be generated to indicate the interaction between people in the group. It can be easily used to be loaded by Gephi, which is a network visualization software, and then generate a network like the following,

NEXT

People may use nicknames a lot during easy chat, especially for classmate groups. wechatAnalyzer will provide a way to load nickname set.