Chatbot Development for Voice Recognition using an Improved Support Vector Classifier

This paper proposes the development of an interactive voice chatbot using Machine Learning (ML) and automation technology. The chatbot will be designed to sense the user's spoken input, process the input using an already trained machine learning algorithm, and generate an appropriate response based on classification. User queries from a given list of typical questions and their respective answers are being labeled using an Improved Support Vector Classifier with TF-IDF vectorization. The trained model is preserved for recall purposes, hence ensuring response creation becomes efficient. For voice interaction support, the speech_recognition library takes the user's voice and converts it into text. The system then executes the text input and creates a sufficient output that it turns into voice with pyttsx3 in a silky-smooth voice interaction. The chatbot also makes use of Selenium WebDriver to perform activities such as opening YouTube or searching for data on Wikipedia, demonstrating more ability than its simple chats. The entire conversation is preserved in a database, and users are allowed access to previous questions using a web interface built with HTML templates. This bot is a demonstration of the potential that ML, natural language processing, and automation can provide in making a smart virtual assistant possible that can be voice-controlled and web-automated and features an interactive and dynamic user interface.