Persian Physical Activity Recommender System Based on ParsBERT

Engaging in regular exercise is essential for maintaining health and preventing various chronic diseases. However, many individuals fail to meet the recommended levels of exercise due to barriers such as a lack of time, motivation, or access to facilities. Consequently, developing personalized and effective interventions to encourage exercise presents a significant challenge. In this paper, we propose a novel Persian exercise recommender system based on ParsBERT, a state-of-the-art language model for Persian natural language processing. The proposed system aims to provide tailored and timely suggestions for exercise to users based on their preferences. It utilizes ParsBERT to understand users' natural language inputs and recommend suitable sports videos. Additionally, it leverages ParsBERT to learn from users' behaviors and preferences over time, adapting its recommendations accordingly. We evaluate our system using a dataset of 100 Persian-speaking users who interacted with the platform. We measure its effectiveness in terms of adherence to exercise recommendations. Furthermore, we compare it with several baseline systems that do not utilize ParsBERT. The results indicate that our system outperforms the baseline systems across all metrics, demonstrating the potential of using ParsBERT to develop intelligent and personalized exercise recommender systems for Persian speakers.