Electronic Thesis and Dissertation Repository


Master of Science


Computer Science


Mike Katchabaw


The concept of believable agent has a long history in Artificial Intelligence. It has applicability in multiple fields, particularly video games. Video games have shown tremendous technological advancement in several areas such as graphics and music; however, techniques used to simulate dialogue are still quite outdated. In this thesis, a method is proposed to allow a human player to interact with non-player characters using natural-language input. By using various techniques of modern Artificial Intelligence such as information retrieval and sentiment analysis, non-player characters have the capability of engaging in dynamic dialogue: they can answer questions, ask questions, remember events, and more. This conversation system is highly customizable, so the types of responses that non-player characters give can be modified to fit within a game’s storyline. Although the system only currently allows for simple dialogue, it illustrates the potential for a more robust way to simulate believable agents in video games.