LAUSR

Films are a high-risk industry. Accurate prediction of movie box-office revenues can reduce this market risk and inform the investment decisions regarding promotion of the movie closer to a film’s release or right after release. Studies have shown that chatter on social media platforms like Twitter along with certain movie-related factors can be useful in predicting success of movies. Sentiment of tweets for any movie gives important information about the consumer’s reaction and the polarity of these sentiments has been shown to have an impact on prediction of box-office revenues. This paper presented a novel Bollywood domain specific sentiment lexicon that delivered state-of-the-art performance for polarity determination of reviews. SentiDraw lexicon was built on movie reviews scraped from IMDB and calculated the sentiment orientation of these words by calculating the probability distribution of words across reviews with different star ratings. The results showed that SentiDraw lexicon delivered a superior performance compared to any other lexicon-based method. This significantly contributed in enhancing the prediction accuracy of box office for movies using textual data from Twitter for analysis. In fact, this study demonstrated an extremely parsimonious regression model that used only budget, hype factor, tweet volume, and polarity of tweets for a robust prediction of box office revenues even before the release of a movie.