##NTHU 2018 Spring Semester Data Science Final Project

Goal

• Using LSTM based on Keras to predict whether the comment in the article is useful

Step

• Crawl the articles from 2018.1 to 2018.3 in PTT Gossiping
• Design a GUI for labeling some comment data
• Using the library 'Jieba' to segement all dataset
• Divide all data into three parts including training, validate, and testing data
• Construce a LSTM model by using training data
• Using testing data to evalute the model

Result

Given a URL from PTT Gossiping, it generates a new HTML file. In the new HTML, it changes the color of the useless comments. We give each user who is in our dataset a credit score. Finally, we also judge whether the article is shit posting, if yes, it will be given a "Shitposting Mark.