Transformer Language Models for UX Dimensions in VR Gaming

This repository contains seven fine-tuned transformer models for analyzing UX dimensions in virtual reality (VR) gaming. The models are based on RoBERTa-large [1] and were fine-tuned on 2,100 VR game reviews to classify sentiment across seven UX dimensions: Immersion, Presence, Achievement, Engagement, Embodiment, Physical Discomfort, and Controllability. Each model assigns one of three possible sentiment labels to user reviews: 0 (neutral): The UX dimension is mentioned neutrally or not discussed. 1 (positive): The UX dimension is described positively. 2 (negative): The UX dimension is described negatively. These models were developed as part of the research study: Pasch, S., Lee, S., Cha, M.C. (2025). Exploring UX Dimensions and Experience-Levels in VR Gaming: Insights from Online Reviews using Transformer Models. Citation: Please cite the corresponding research paper when using these models in your work. Additional References: [1] Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., ... & Stoyanov, V. (2019). RoBERTa: A robustly optimized BERT pretraining approach. arXiv preprint arXiv:1907.11692.本仓库包含7个经微调的Transformer模型，用于分析虚拟现实（Virtual Reality，VR）游戏中的用户体验（User Experience，UX）维度。上述模型基于RoBERTa-large[1]开发，在2100篇VR游戏评测数据集上完成微调，旨在对7个UX维度的情感倾向进行分类，涵盖沉浸感（Immersion）、临场感（Presence）、成就感（Achievement）、参与度（Engagement）、具身感（Embodiment）、身体不适（Physical Discomfort）与可控性（Controllability）。 每个模型可为用户评测分配三类情感标签之一：0（中性）：对应UX维度被中性提及，或未被相关讨论；1（正面）：对应UX维度被正面描述；2（负面）：对应UX维度被负面描述。 本系列模型作为下述研究工作的组成部分开发完成：Pasch, S., Lee, S., Cha, M.C. (2025). 《探索VR游戏中的用户体验维度与体验层级：基于Transformer模型的在线评测分析》。 引用说明：若在研究工作中使用本模型，请引用上述相关研究论文。 附加参考文献：[1] Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., 等 & Stoyanov, V. (2019). RoBERTa：一种经鲁棒优化的BERT预训练方法。arXiv预印本arXiv:1907.11692。