Izvorni znanstveni članak
https://doi.org/10.32985/ijeces.17.7.4
A Bilingual Sentiment Analysis Platform for Medication Reviews: A Transformer-Based System Supporting Active Learning and Ongoing Corpus Expansion
Fatima Zohra Youcef
; University of Oran 1 Ahmed Ben Bella Laboratoire d’Informatique d’Oran, Department of Computer Science Oran, Algeria
*
Fatiha Barigou
* Dopisni autor.
Sažetak
In contemporary healthcare systems, patient feedback constitutes a valuable source of information for understanding real- world medication use and treatment effectiveness. Sentiment analysis contributes to pharmacovigilance by facilitating the detection of adverse effects and strengthening drug safety monitoring through the systematic interpretation of patient-reported experiences. However, several persistent challenges restrict its full potential. The lack of comprehensive platforms for analyzing patient-reported outcomes limits healthcare providers’ ability to derive meaningful insights into medication effectiveness. Furthermore, most existing research remains predominantly in English, while other important languages—particularly Arabic—receive comparatively limited attention. This imbalance is exacerbated by the scarcity of Arabic-language resources and annotated datasets, which continues to hinder the development of inclusive, linguistically diverse healthcare solutions. To address these gaps, we introduce BI-PharmaP, a bilingual platform for analyzing medication reviews in both English and Arabic. The platform is structured around four key components: (i) a unified interface that enables patients and healthcare providers to query and analyze existing feedback, explore comments and statistical visualizations, and submit new reviews; (ii) a bilingual processing strategy that ensures consistent handling of English and Arabic inputs, including dialectal variations; (iii) a fine-tuned RoBERTa model for sentiment classification; and (iv) an active learning loop that incorporates user-validated corrections to iteratively refine the model’s performance. In addition, the platform facilitates the continuous expansion of a bilingual labeled review corpus, thereby supporting future research and model development. Collectively, BI-PharmaP provides a scalable and adaptive framework that enhances pharmacovigilance, strengthens drug safety monitoring, and advances patient-centered pharmaceutical research.
Ključne riječi
Bilingual sentiment analysis; transformers; Active Learning; Arabic corpus expansion; user feedback; pharmacovigilance; medication reviews; Sentiment Analysis platform;
Hrčak ID:
348738
URI
Datum izdavanja:
1.7.2026.
Posjeta: 0 *