Estimation of Reader's Emotion to News Articles Using Twitter and its Application to Access Analysis

概要

論文の詳細を見る
In this paper, we propose an estimation of reader's emotion towards news articles using Twitter, and apply this method to the access log analysis of a news website. The proposed estimation method is divided into 3 steps: 1) The system collects tweets about news articles from Twitter. 2) The system tags each tweet with an emotion. 3) The system estimates the reader's emotion towards the news articles using tweets obtained in the second step. The system can tag an article with one or more of 11 emotions (anger, dislike, excitement, fear, like, joy, relief, sadness, shame, surprise, nothing). The experimental result shows the effectiveness of this method using Twitter for estimating the reader's emotion. For access log analysis based on emotion, we collected the access log of a self-produced news website. The system tags each collected news article with emotions using the proposed method. We analyzed the access log by using Hidden Markov model (HMM). As the result of access log analysis, we obtained some interesting findings such as that the users tend to read the article which causes excitement or surprise at the start.