Function

• Crawl comments under zhihu answer, and form one round conversation , the answers are from collution"赞同超过1000的回答
• Assign the emotion to each sentence in QA pairs using NLPCC emotion labeled data of weibo

Sample Data

the data is stored in the json format. it is a dictionary list, contains three field: "post" ,"res" for response, and the conversation id.

Requirement

• zhihu_outh
• torch
• pandas
• jieba
• numpy

Usage

• fill in you zhihu account in login.py and generate token.pkl.
• run crawler.py to generate the conversation and run preprocessing.py to clean and format data.
• download NLPCC 2017 weibo data and place it in ./data folder.
• run Classify/cnn_pytorch.py to generate the emtion label for the sentences.