An Intelligent Medicine Recommendation System Using NLP, BERT, and Medical Knowledge Graph

Abstract

Healthcare systems generate large volumes of unstructured data such as patient reviews, prescriptions, and clinical notes, making it challenging to extract meaningful insights for decision-making. This paper proposes a medicine recommendation system using Natural Language Processing (NLP) combined with advanced deep learning techniques to improve accuracy and reliability. The system utilizes Bidirectional Encoder Representations from Transformers (BERT) to perform context-aware sentiment analysis of patient reviews, enabling better understanding of drug effectiveness and side effects. Additionally, a medical knowledge graph is integrated to ensure clinically safe recommendations by validating drug–disease relationships and identifying contraindications.
To enhance usability, the system incorporates personalization based on patient-specific factors such as age, medical history, and allergies. The proposed model follows a structured pipeline including data preprocessing, feature extraction, sentiment analysis, safety validation, and recommendation generation. Experimental evaluation demonstrates that the system outperforms traditional machine learning approaches in terms of accuracy, precision, and recommendation quality.
The proposed approach provides a reliable, efficient, and scalable solution for intelligent medicine recommendation, with potential applications in telemedicine and digital healthcare systems.