Machine Learning and Speech Analysis Framework for Protecting Children against Harmful Online Content

Nowadays, it becomes significant to protect children from the detrimental influences of offensive online things like hateful messages, violence content, bullying and pornography that are now broadly spreading on Internet. Thus, Machine Learning (ML) algorithms can diagnose such content and then remove them. This paper depicts a unique ML model to detect the abusive content via speech recognition technique, especially for children. ML provides efficient tools for the assessment of web content and filtering of offensive material. The performance outcomes of developed model validate its accuracy and effectiveness. With the advancement of several methodologies, machine-learning approach proves its dependency on the quality of information that involves diversity and representativeness. ML can only handle such dataset efficiently if the developed algorithm has the potential to deal with poor-quality training datasets. The human moderation experts needed to interpret the contextually hard content as per the current scenario or even in foreseeable future in spite of the progress in ML technique. The key research activities under online content domain using ML approaches because of the massive volume of information for researching/training models. This developed ML algorithm also shows its effectiveness to deliver accurate outputs for the cases of limited training data or experiencing issues in understanding them.