Wechat Kuakua chatbot

Wechat is the dominant social networking platform in China. This project is inspired by the trend that people gives compliments to each other in wechat group, which means "kuakua" in Chinese.

itchat: auto reply chatbot using tuling api

The auto-reply chat bot use tuling API, Apply a tuling chatbot and get a API url from the website, such as

apiUrl = 'http://www.tuling123.com/openapi/api'

Note down the key when apply and then get the auto-response from the API

data = {'key': KEY, 'info': msg, 'userid' :'wechat-robot',}    
try:        
	r = requests.post(apiUrl, data=data).json()        
	return r.get('text')    
except:        
	return 

Then itchat would help reply the response back from the API to Wechat using the following line before the function header.

@itchat.msg_register(itchat.content.TEXT)

Use followig code to run itchat.

itchat.auto_login(enableCmdQR=2, hotReload=True)
itchat.run()

Wechat_kuakua1.0

Reference: A Wechat Article

wechat_kuakua1.0 uses a dict of corresponding keywords with an array of prepared replies. The program then matches words appearing in the message sent by the user with the keywords and replies randomly in the corresponding array of prepared replies. )

REOLY = {'keyword': ['sentence 1', 
                    'sentence 2'],
        'keyword2': ['sentence 3',
                    'sentence 4']}

@itchat.msg_register([TEXT], isGroupChat=True)
def text_reply(msg):
    group_name = 'group name' # put wechat group name here
	match = re.search('keyword', msg['Text'])
    if match:
        randomIdx = random.randint(0, len(REPLY['keyword']) - 1)
        itchat.send('@' + '%s\n%s' % (username, REPLY['keyword'][randomIdx]), msg['FromUserName'])

Note: set a wechat group, and set its name variable group_name and all the conversation would only take place in that wechat_group.

Using Chinese Nlp Library

Reference

The corpus is crawled from douban. We parse the sentences of messages and train the model use Genism.

First load the question and answer from the corpus into an array of messages, and an corresponding array of lists of replies. The index is used to connect the message and its replies.

Then we parse the received message and compared it to the sentences of messages we have in the corpus. The similarity matrix can be expressed by models TF-IDF, LSI, LDA etc. We search for the top 4 similar sentence and choose a random one, reply by its list of replies in the corpus.

Run

Tp run the program, cd into the directory and run python3, such as

python3 wechat_kuakua2.0.py
python3 wechat_kuakua1.0.py

Need to install and import

• itchat, an easy-to-use personal wechat API that enables wechat to auto-apply messages,
• jieba, a Chinese word segmentation module, and
• Genism, a Python library for topic modelling, document indexing and similarity retrieval with large corpora. Recommend virtualenv to develop the program.

A useful article to understand how genism works.