Prediction of Oscar award nominations based on movie scripts.

摘要

According to www.the-numbers.com/market/, gross revenue for the Hollywood movie industry was over USD 10 billion in 2009. With annual revenues at this scale, it is critical for a movie to be successful at the box office. High revenue is closely linked to Oscar award nominations and hence to winning of the award. Building predictive models for nominations can provide useful insights into predicting Oscar awards. Using a movie script as a movie representation, we rank movie scenes and retrieve features from regression of the movie scene ranking. We introduce a model using Support Vector Machines, which predicts Oscar award nominations in the Best Screenplay and Best Picture categories. We build three baselines using random decision, bag-of-words and IMDb features respectively. Prediction results show that features from movie scripts contain information which is useful in predicting Oscar award nominations.