« Back to blog

"Film Forecaster", a tool to help predict the box office from Twitter data

I just read in Gonzalo Martin's blog that the USC Annemberg Innovation Lab has released a tool to analyze the amount and quality of the tweets (positive or negative) related to a movie release during a weekend. You can see the example from this weekend here. Gonzalo asks if looking at the first samples, some communication actions could be done to reconduct negative buzz and promote the theaters attendance at the last moment, thanks to these kind of tools and geolocalization. It seems to me a tool with a really interesting potential, and I can imagine a scenario like in Truman's Show, where all movie marketers will be following social networks during all the weekend, to see exacly what are our Trumans, sorry, our potential audience, saying and doing. It has an evil touch, but I have no doubts that it could be a key tool to understand why successes and failures happen, and also, like Gonzalo says, to live reconducting little buzz crisis. However, I ask myself how tweets are classified, in an automated manner, between beign "positive" or "negative" buzz of the movie. With key words? Do they consider sarcasm and irony? What about neutral informations? It would be interesting too to obtain information about the causes of this positive or negative buzz: if we detect that there's a lot of negative buzz about our movie release, the very next thing we are going to need in order to reconduct the situation is to know why. It may help some "most used words" analysis in each category? How could we do this qualitative analysis in a massive and automated way? 

Jjjjj