Development of novel classifying system to identify the right sense of audio conversation in social networks using deep convolution neural network

Abstract

Social media has paved a new way for communication and interacting with others. The use of social media differs according to the socio-cultural, demographic and psychological aspects of individuals. People chat, share ideas and visual material, and feel that they satisfy their needs of belonging along with the groups they have joined. Social networks is not only a area of freedom where persons express themselves openly or furtively, but also an area where several ways of violence emerge or even a means used for some aspects of violence.. The present research throws light on a few of the regular and trendy methods of abuse and risks faced by the users of social media. Develop a system to identify abusing audio file by an individual on a people/ group based on common language, race, sexual preferences, religion, or nationality. We examine a new model from machine learning, namely deep machine learning by probing design configurations of deep Convolutional Neural Networks (CNN) and the impact of different hyper-parameter settings in identifying the negative aspects in social media. Deep CNN automatically generate powerful features by hierarchical learning strategies from massive amounts of training data with a minimum of human interaction or expert process knowledge. An application of the proposed method demonstrates excellent results with low false alarm rates for Twitter data.