學術資源整合系統-相關推薦
 |
|
|
|
|
| 作者 | Chou, Yin-Yu;Huang, Chun-Yen;Lin, Yi-Ru;Zhang, Rong-Han;Hsu, Chia-Yu;Chen, Ying-Ting |
|---|
| 出版日期 | 20260204 |
|---|
| 著作名稱 | Integrating Large Language Models into Coffee Shop Simulation Games: NPC Dialogue Generation and Educational Applications
|
|---|
| 會議名稱 | The 10th International Conference on Machine Learning and Soft Computing |
|---|
| 會議地點 | Osaka, Japan |
|---|
| 主辦單位 | ICMLSC 2026 |
|---|
| 國際性會議 | Y |
|---|
| 其他資訊 | EI |
|---|
| 主題 | 資訊科學;科學教育 |
|---|
| 關鍵字 | Large Language Model (LLM), Non-Player Character (NPC), Simulation Game, Human–AI Interaction. |
|---|
| 摘要 | This study presents the design and implementation of a coffee shop
simulation game that integrates large language models (LLMs) to enhance non-
player character (NPC) interactions. Unlike conventional management games
that rely on fixed-text dialogues and simplified workflows, this system employs
Google Gemini to generate dynamic, personality-driven NPC conversations.
The game, developed with Unity, includes multiple modules such as coffee
preparation, customer service, knowledge-based mini-games, a “Dream Tree”
progression system, and a sustainability-oriented greenhouse. Two dialogue
generation modes were implemented: a menu-driven mode ensuring recipe
accuracy and a free-form mode simulating unpredictable customer requests.
Evaluation focuses on dialogue quality, system latency, player immersion,
learning outcomes, and sustainability awareness. Preliminary playtests suggest
that Gemini-powered NPCs improve realism, enhance interactivity, and support
knowledge acquisition. The study highlights both contributions and challenges:
while LLM integration enriches immersion, education, and innovation in game-
based learning, issues such as content control, computational cost, and large-
scale validation remain. These findings demonstrate the potential of LLM-
driven NPCs in advancing intelligent computing applications and educational
game design. |
|---|
| 資料連結 |  |
|---|
| 系統號 | NO000007736 |
|---|
|
|