Predicting Quality of Movie from Metadata and Plot Summary

Abstract

The worldwide theatrical market had a box office of US $42.2 billion in 2019. In recent years it has been seen that it is growing even more and more, as a consequence urge to predict the success of the movie has increased. To inspect this issue various methodology has been proposed some of which rely on reviews and the trailer when most or all the budget of the movie has been enervated. To overcome this some recent papers have also used the plot summary of the movie to classify the movie as successful or not successful. In this work, we will try to predict the quality of the movie not only by using the plot summary but other metadata of the movie too. We have used the CMU corpus for the movie metadata and the IMDB database for the ratings. We have experimented with LSTM, ELMO, Sentiment analysis, and Transformer based architecture like BERT. We have experimented with all these and combined them to come up with a feature engineering architecture suitable for our task.