Disturbance Detection in Mobile Instant Messaging Apps using Deep Learning

This paper demonstrates the design of a Mobile Instant Messaging application with a feature to filter distracting content, thereby increasing productivity at work. Handling a hugeinflux of incoming texts is a frustrating task for one already burdened with work. Therefore, an option can be provided for filtering content, as shown in this paper. From the literatures, it was observed that there is no existing technology to solve the aforementioned problem. It has been addressed in this paper by means of a deep learning architecture capable of classifying text and images into important and distracting categories. Google BERT model was used for classification of text and Resnet-34 was used for image classification; accuracies of 98.065% and 99.34%were obtained for these models respectively.