Machine-Learning-Based Profanity Detection and Removal in Cartoons Videos

Swear words and offensive languages in cartoons are prohibited and should be avoided by our children. This paper aims to detect swear words and remove them from cartoons videos for the kids. The system was implemented via a web-based application. The user is allowed to upload cartoons videos to the cloud and footage of video is converted to text using Google Speech-to-Text. We used the linear SVM to detect all swear words which appeared in a text file. The system is capable of muting these swear words to the audio file. After the removal of swear words, the result of cartoons videos can be downloaded and playback offline. The precision, recall, F1-score, and accuracy of our system are 0.957, 0.958, 0.931, and 0.958 respectively.