Enhancing AI-based Conversational Dialogue Systems

Dialogue system is a popular Human-Computer Interaction (HCI) approach in daily life. The advancement of natural language processing (NLP) has made dialogue systems more realistic and commercially viable (e.g., moving booking chatbot). This thesis aims to improve AI-based dialogue systems by these AI-related methodologies. This thesis proposes 1) a dialogue system to extract both similarity and semantic dependencies in conversations; 2) a rich text conversational dataset extracted from cybersecurity domain; 3) a novel attention mechanism to conduct better word representations.