Cameron Main


← Back to Projects

Dissertation Project


Evaluating NPC Dialogue Interaction Methods in Video Games.

Project Overview


This research project aimed to evaluate the effectiveness of different NPC dialogue interaction methods in video games. The project incorporated a novel method of interaction using NLP and LLM AI tools and technology within Unity. I successfully implemented conventional NPC dialogue interaction methods, conducted user research, and explored open & closed domain dialogue generation systems.

Dissertation Project Screenshot

Key Features


Technology


C# Unity Python spaCy NLP OpenAI ChatGPT API

Achievements


The project successfully incorporated both conventional and experimental NPC dialogue interaction methods. The branching dialogue system allowed for interactive and multi-path conversations, while the NLP and LLM-based methods provided players with a heightened degree of agency and immersion. User research and playtesting feedback added valuable insights into how gamers perceive NPC interactions within games.

Challenges and Improvements


As a solo developer, meeting deadlines and incorporating all requirements within the limited development time was challenging. The effectiveness of the LLM interaction method could have been enhanced with more time. In hindsight, prioritizing the more complex experimental interaction methods and allocating more time for their development would have been beneficial.

Future Development


The project highlighted the potential of using NLP tools and techniques in NPC interaction but also emphasised the need for further development and refinement. Future work should focus on enhancing the LLM method to address issues like NPC repetition and unnatural conversation breakdowns. However, the ongoing importance of branching dialogue systems in video games is recognised, and their continued use for stable and reliable NPC interactions is recommended.

Read the Paper