Combining structure, content and meaning in online social networks: The analysis of public's early reaction in social media to newly launched movies

• Published in: Technological forecasting & social change Vol. 109; pp. 35 - 49
• Main Authors: Lipizzi, Carlo, Iandoli, Luca, Marquez, José Emmanuel Ramirez
• Format: Journal Article
• Language: English
• Published: New York Elsevier Inc 01.08.2016; Elsevier Science Ltd
• Subjects: Broadcasting; Conversation; Data mining; Diffusion; Digital media; Dynamic tests; Information dissemination; Internet; Mass media; Mass media images; Mathematical models; Motion pictures; Networks; Online; Power; Predictions; Predictive modeling; Sentiment analysis; Social behavior analysis; Social media; Social networks; Studies; Technological forecasting; Text mining; Traffic; Twitter case study; Text mining; Twitter case study; Predictive modeling; Social behavior analysis; Data mining; Social media
• ISSN: 0040-1625; 1873-5509
• DOI: 10.1016/j.techfore.2016.05.013

In this paper we present a methodology to assess moviegoers' early reactions to movies' premieres through the extraction of analytics from Twitter conversations that take place in the weekend in which a movie is released. We then apply data mining techniques to a sample of 22 movies to identify models able to predict box-office sales in the first weekend. We show that better predictions are obtained when traffic metrics are combined with social network or conversational indicators rather than with sentiment and that online sentiment achieves the lowest explanatory power among all the considered variables. Our findings confirm that the importance of commonly used buzz-metrics, such as sentiment, is probably overstated, and that conversational analytics can contribute significantly to explain the variance of box office revenues in the first week end of release. More broadly, our work adds to research on information diffusion in online networks by providing evidence that diffusion of messages is not content-neutral and that the analysis of conversational dynamics can help to understand the interplay between collective generation and diffusion of content in social networks as well as to obtain insights on whether information diffusion influences off-line behavior. •Moviegoers' reactions to movies' premieres to predict success of movies using social media streams•Streams are modeled as virtual conversations, from which we extract dynamic concept maps.•We use data mining to predict movies' success from semantic, topology, traffic and opinion metrics.•The best prediction is obtained combining traffic metrics, social network and semantic metrics.•Dominant buzz-metrics based on sentiment and social network analysis may be overstated.