<![CDATA[This study aims to develop a visual novel game in which the user can actively interact with an NPC by entering their own response rather than selecting dialogue options, as well as to improve the immersion of the Player-NPC dialogue system by obtaining the player's in-game text inputs while using intent recognition with the BERT transformer model. The researchers developed two different versions of the visual novel game to achieve their goal. The first version of the game uses the standard visual novel game mechanics, while the second version uses text input to communicate with NPCs, and the BERT algorithm was integrated in this version for the intent recognition of the player's text inputs. The retraining of the model achieved an accuracy of 0.9375. To measure the players’ immersion in both versions, a Modified Immersion Experience Questionnaire was given to the players to complete after each playthrough of the visual novel games. Based on the research findings, the researchers concluded that when participants played the game that uses Intent Recognition with BERT, their level of immersion significantly improved.]]>
Copyrights © 2023
