Design of a pilot study to evaluate a Question Answering model based on BERT

The increasingly widespread use of technologies to support teaching o↵ers the sci- entific community new scenarios in which they can apply and combine technolo- gies in innovative ways, always looking to improve teaching techniques. Among these technologies, an emerging trend is the adoption of Artificial Intelligence (AI) systems combined with Extended Reality (XR) technologies. In education, the combination and correct usage of these technologies can boost the learning experience of students, leading to the acquisition of new knowledge and enhanc- ing their skills, such as computational thinking and creativity [2–4]. One tangible example of the combination of these technologies is presented in Farella et al. [1]. The paper proposes a Virtual Reality (VR) system composed of a 3D avatar trained to answer questions posed by the user, using the BERT language model. In the abstract we present here, a pilot study of the model described in Farella et al. [1] is proposed, to evaluate the quality of the trained question-answer model and the correct functionality of the system.