Word of Mouth in Online Social Network, Using Twitter to Predict Box Office Revenue

In this paper, we study the economic impact of user generated contents (UGC) in twitter on movies’ box offices. In contrast to other UGC such as online reviews, twitter more truthfully documents a product’s word of mouth (WOM) among ordinary consumers. We develop a new text mining technique to extract and analyze the movie conversations among twitter users. This mining technique is specially-developed for twitter texts and is unsupervised in nature. The core of the technique is an opinion miner that analyzes the contents and sentiments of twitter messages, with a very competitive accuracy consistently exceeding 75% for various tasks. This text mining technique enables us to extract user’s overall opinions automatically in large scale. We further drill down the actual contents of twitter messages by analyzing the sentiments of a message with respect to the four main attributes of a movie: plot, acting, directing and visual effects. Lastly, we propose a novel and important metric – disposition to capture a blogger’s prior opinion toward a random movie. The outputs from twitter mining are then used to predict the box office of movies to understand the power and economic impact of twitter conversations.