Using personality parameters in simulating dialogue text for converstional agents

• Main Author: Alicbusan, Roscoe Nealle Chua
• Format: text
• Language: English
• Published: Animo Repository 2018
• Subjects: Computer-assisted instruction > Programming; Programmed instruction; Teaching > Aids and devices
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/5530
• Institution: De La Salle University

Numerous researches have been conducted with the purpose of creating more human-like agents. A popular type of agent in this endeavor is conversational agents or chatbots, which aim to be able to hold a conversation with human users. Human-like traits such as personality, mood, and context are integrated to these agents. While there have been strides in incorporating these to conversational agents, the lack of lexical and stylistic features of speech that can highlight the personality of an agent remains an open challenge. Existing conversational agents have personalities manifest themselves differently such as in behavioral patterns, facial expressions, and gestures. This research aims to develop a system wherein users interact with conversational agents which generate context-sensitive dialogue influenced by personality-related lexical and stylistic features. Tests with children from between 7 to 12 years old show that only certain noticeable features were apparent to the users during the conversation while the character's behaviors were more easier for them to discern. Furthermore, the character's behavior affects the user's perception of said character and even their enjoyment in the conversation as a whole. On the other hand, expert analysis show that some characters were having unexpected responses and that the conversations themselves are very unnatural largely due to the abundance of Yes or No questions.