Nova: A Machine Learning–Enhanced Virtual Personal Assistant for Windows with Voice and Text Interaction

Abstract - Switching between many applications to check the weather, search for information, play media or send a message makes everyday computer use slower than it needs to be. This paper presents Nova, a Python-based virtual personal assistant for Windows that lets users interact with their computer through voice and text commands. Nova can report the weather, search Wikipedia, play YouTube content, set reminders, perform calculations, take screenshots, open applications and websites, and help with e-mail and WhatsApp messages. To understand commands phrased in different ways, a Machine Learning layer is added to the Natural Language Understanding stage, using TF-IDF features and a Logistic Regression classifier trained on 14 predefined intents. When the prediction is not confident enough, a rule-based router handles the command instead. The assistant is built in layers, with a Tkinter interface, a control layer, task modules, external service integrations and a speech-based response layer. Nova combines voice interaction, language processing, Machine Learning and task automation into a simple, lightweight and extensible desktop assistant.