Predicting the Popularity of Topics based on User Sentiment in Microblogging Websites

TL;DR

This paper explores how community sentiment energy, derived from user emotions on key phrases, correlates with and can predict the popularity of topics in microblogging communities using models like MRF and graph entropy.

Contribution

It introduces a novel approach to quantify community sentiment energy and demonstrates its effectiveness in predicting topic popularity in social media.

Findings

Strong linear correlation between sentiment energy and topic popularity

Proposed models effectively predict topic spreading based on sentiment

Community sentiment influences collective decision-making in topic dissemination

Abstract

Behavioral economics show us that emotions play an important role in individual behavior and decision-making. Does this also affect collective decision making in a community? Here we investigate whether the community sentiment energy of a topic is related to the spreading popularity of the topic. To compute the community sentiment energy of a topic, we first analyze the sentiment of a user on the key phrases of the topic based on the recent tweets of the user. Then we compute the total sentiment energy of all users in the community on the topic based on the Markov Random Field (MRF) model and graph entropy model. Experiments on two communities find the linear correlation between the community sentiment energy and the real spreading popularity of topics. Based on the finding, we proposed two models to predict the popularity of topics. Experimental results show the effectiveness of the two models and the helpful of sentiment in predicting the popularity of topics. Experiments also show that community sentiment affects collective decision making of spreading a topic or not in the community.