The use of Artificial Intelligence (AI) in education is growing rapidly, transforming the teaching-learning process as well as the assessment process.
This work introduces SLASys, a tool to recommend the assessment of short-answer questions using semantic AI techniques. Unlike other works based on generative AI, SLASys uses the lightweight BERT language model, which better understands specific domain language concepts, improves computational efficiency, and reduces ethical and privacy concerns.
SLASys implements semantic comparison and predictive classification models based on BERT. A mixed research methodology was followed, combining action research with a design and creation approach, to develop and refine SLASys over four editions of a master's-level course on patent examination within the intellectual property domain.
SLASys has been integrated into Moodle, enabling its use by teachers without technical expertise, and has been tested by 120 students.
The results demonstrate its effectiveness even with small datasets and limited participants within the described experience and according to existing literature. Additionally, it has been positively evaluated by both teachers and students.
This work shows the feasibility of using AI in higher education, in both hybrid and online environments, offering a practical solution to improve assessment and feedback for short-answer questions in real learning contexts.
---
How to Cite: Bañeres Besora, D., Guerrerro Roldán, A.-E., & Rodríguez González, M. E. (2026). A language model-based recommender assessment system for short-answer questions in the intellectual property domain. RIED-Revista Iberoamericana de Educación a Distancia, 29(1), 321–352. https://doi.org/10.5944/ried.45541
