MPAA Rating Prediction Based on Deep Learning

The Motion Picture Association of America(MPAA) is used in the United States to rate a film's compatibility via its content. These systems evaluate ratings according to the scenes in the movies. In this work, a deep learning based classification technique is proposed to differentiate between MPAA ratings (G, PG, PG-13, R, NC-17 ) via subtitles. Movie subtitles which generally locate bottom of the screen to show character dialogs in movies or series. The syntax of each subtitle files are learned through Word2Vector. We then constructed a binary classifier based on the preceding representation dataset. We studied the performance of different classifiers with stratified ten-fold cross-validation. The model has been validated with experiments on English subtitle dataset. According to the experiment results, our proposed method was achieved more than 59\% accuracy rate.