Development of an on-device deep learning textual analytics application using the Android Platform

Ramón Hernández García. (2022). Development of an on-device deep learning textual analytics application using the Android Platform. Final Career Project (TFG). Universidad Politécnica de Madrid, ETSI Telecomunicación.

Abstract:
In the last decades the Internet has seen an exponential growth in the number and impor- tance of the social networks. Such is the case that each organization behind them has to establish some rules for its correct use and hire moderators to check the published content and to limit or block those messages that could damage some people or incite violence. This problem is bigger if we take into account that these online services can be used by all ages citizens. Therefore, people with different knowledge of the Internet can use them. And since the network interface sometimes lead to think it is anonymous, people are more encouraged to avoid taking a moment to think about if what they will publish could have bad consequences. Because of this reason, in this work an Android application will be developed to address this issue. This application will allow users to analyze their messages before sending them, allowing the user to know if it can be offensive. As WhatsApp is one of the most extended networks in the world, the application will target that network. Messages will be analyzed with Machine Learning models of Natural Language Pro- cessing (NLP). For that task, the TensorFlow library developed by Google will be used, and its derived framework TensorFlow Lite for low-end devices with less capable hardware resources, as Android devices can be.