Case Study: Health-Links achieves accurate, aspect-based native Arabic sentiment analysis at scale with Repustate

Arabic Sentiment Analysis Solution for Health-Links

Health-Links, a Jeddah-based healthcare consultancy working with the Saudi Ministry of Health and Press Ganey, needed to analyze over 12 million annual patient and employee verbatim survey responses—mostly in Arabic and often mixed with English—without manual bias, and to report aspect-level sentiment aligned to the Saudi Complaints Taxonomy. To meet these requirements, Health-Links engaged Repustate for a cloud-based, native Arabic sentiment analysis solution capable of automated, aspect-based classification.

Repustate implemented a custom, native-Arabic, aspect-based sentiment model (extended to include newly discovered topics), able to classify comments as Positive, Negative, Neutral, or Mixed and to process hundreds of API calls in seconds. After three iterations within a month the model reached 81% accuracy, enabled analysis of historical and current data at scale, reduced human error, and provided Health-Links and the Ministry with granular, actionable sentiment scores to prioritize policy decisions and improve patient care.

Health-Links

Adel Al-Shabaan

Managing Director