KULLANICI YORUMLARIYLA İLAÇ YAN ETKİSİ TESPİTİ: TRANSFORMATÖR TABANLI DUYGU ANALİZİ ÇIKARIMI

Yıl 2025, Cilt: 24 Sayı: 48, 414 - 437, 18.12.2025

Şeyda Karcı , Sevim Ceylan Böcekçi , Sadettin Demir , Elif Özceylan , Ayşe Berna Altınel Girgin

https://doi.org/10.55071/ticaretfbd.1635145

Öz

Sağlık alanında, özellikle ilaç incelemeleriyle ilgili olarak kullanıcı tarafından oluşturulan içeriğin artan hacmi, olumsuz ilaç reaksiyonlarını (ADR'ler) ve kullanıcı duygularını analiz etmek için önemli fırsatlar sunmaktadır. Bu çalışmada, duygu analizi yapmak için ilaç yan etkileriyle ilgili kullanıcı yorumlarını içeren birleştirilmiş CADEC+ADE veri kümesini kullanıyoruz. Bu çalışma, bir hasta perspektifinden ilaçların etkisini daha iyi anlamak amacıyla kullanıcı geri bildirimlerinin olumlu ve olumsuz duygularının ‘yan etkisi var’ ya da ‘yan etkisi yok’ şeklinde ikili bir sınıflandırmasını içerir. Bunun için beş farklı yöntemin performansını değerlendiriyoruz ve karşılaştırıyoruz: BiLSTM BioBERT, MBERT, SVM ve XLM-R. SVM dışındaki diğer modeller biyomedikal metinler üzerinde önceden eğitilmiş transformatör tabanlı model oldukları ve biyomedikal terminolojiye ilişkin üstün anlayışları için kullanılırken, SVM karşılaştırma için klasik makine öğrenimi temellerini sağlar. Modeller, yan etkilerine ve algılanan etkinliğe odaklanarak, ilaçlarla ilgili kullanıcı tarafından bildirilen deneyimleri içeren birleştirilmiş CADEC+ADE veri kümesinde eğitildi ve değerlendirildi. Sonuçlar BioBERT modelinin %99,2 doğruluk değeriyle bu alan için duygu analizinde diğer tekil yöntemlerden daha etkili olduğunu ortaya koyuyor. Buna ek olarak BERT modelleri ile oluşturulan topluluk öğrenmesi modelinde ise %99,6 oranında doğruluk değeri elde edilmiş olup, bu yöntem ile en yüksek doğruluk oranına ulaşılmıştır.

Anahtar Kelimeler

ADE Corpus , Bert , CADEC , Olumsuz İlaç Reaksiyonu , Transformer Model

ADVERSE DRUG REACTION DETECTION FROM USER COMMENTS: A TRANSFORMER-BASED SENTIMENT ANALYSIS APPROACH

Öz

In the healthcare domain, the increasing volume of user-generated content, especially related to drug reviews, offers significant opportunities to analyze adverse drug reactions (ADRs) and user sentiment. In this study, we utilize the combined CADEC+ ADE dataset, which contains user comments related to adverse drug reactions to perform sentiment analysis. This work aims to better understand the impact of medications from a patient perspective by conducting a binary classification of user feedback into “has adverse drug reaction" or ‘not’ based on positive and negative sentiments. To achieve this, we evaluate and compare the performance of five different methods: BiLSTM, BioBERT, MBERT, SVM, and XLM-R. While SVM provides a classical machine learning baseline for comparison, the other models are transformer-based architectures pre-trained on biomedical texts, chosen for their superior understanding of biomedical terminology. The models were trained and evaluated on the combined CADEC_ADE++ dataset, which includes user-reported experiences related to drugs with a focus on adverse drug reaction and perceived efficacy. The results reveal that the ensemble learning model created with BERT models, an accuracy of 99.6% was obtained and the highest accuracy rate was achieved with this method.

Anahtar Kelimeler

ADE Corpus , Adverse Drug Reaction , Bert , CADEC , Transformer Model

Kaynakça

• Agarwal, A., Sharma, P., Alshehri, M., Mohamed, A. A., & Alfarraj, O. (2021). Classification model for accuracy and intrusion detection using machine learning approach. PeerJ Computer Science, 7, e437.
• Aklouche, B., Bounhas, I., & Slimani, Y. (2018). Query Expansion Based on NLP and Word Embeddings. TREC.
• Al-Juboori, S. A. M., Hazzaa, F., Jabbar, Z. S., Salih, S., & Gheni, H. M. (2023). Man-in-the-middle and denial of service attacks detection using machine learning algorithms. Bulletin of Electrical Engineering and Informatics, 12(1), 418–426.
• Bergman, E., Dürlich, L., Arthurson, V., Sundström, A., Larsson, M., Bhuiyan, S., Jakobsson, A., & Westman, G. (2023). BERT based natural language processing for triage of adverse drug reaction reports shows close to human-level performance. PLOS Digital Health, 2(12), e0000409.
• Chauhan, T., & Palivela, H. (2021). Optimization and improvement of fake news detection using deep learning approaches for societal benefit. International Journal of Information Management Data Insights, 1(2), 100051.
• Chen, D., Bourlard, H., & Thiran, J.-P. (2001). Text identification in complex background using SVM. Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001, 2, II–II.
• Conneau, A. (2019). Unsupervised cross-lingual representation learning at scale. arXiv Preprint arXiv:1911.02116.
• Dong, X., Yu, Z., Cao, W., Shi, Y., & Ma, Q. (2020). A survey on ensemble learning. Frontiers of Computer Science, 14, 241–258.
• Ekong, B., Edet, A., Udonna, U., Uwah, A., & Udoetor, N. (2024). Machine Learning Model for Adverse Drug Reaction Detection Based on Naive Bayes and XGBoost Algorithm. British Journal of Computer, Networking and Information Technology, 7(2), 97–114.
• Eslami, B., Rezaei, Z., Habibzadeh, M., Fouladian, M., & Ebrahimpour-Komleh, H. (2020). Using deep learning methods for discovering associations between drugs and side effects based on topic modeling in social network. Social Network Analysis and Mining, 10, 1–17.
• Guo, D., Wang, Q., Liang, M., Liu, W., & Nie, J. (2019). Molecular cavity topological representation for pattern analysis: A NLP analogy-based Word2Vec method. International Journal of Molecular Sciences, 20(23), 6019.
• Gurulingappa, H., Rajput, A. M., Roberts, A., Fluck, J., Hofmann-Apitius, M., & Toldo, L. (2012). Development of a benchmark corpus to support the automatic extraction of drug-related adverse effects from medical case reports. Journal of Biomedical Informatics, 45(5), 885–892.
• Haq, H. U., Kocaman, V., & Talby, D. (2022). Mining adverse drug reactions from unstructured mediums at scale. In Multimodal AI in healthcare: A paradigm shift in health intelligence (pp. 361–375). Springer.
• Jakkula, V. (2006). Tutorial on support vector machine (svm). School of EECS, Washington State University, 37(2.5), 3.
• Karcioğlu, A. A., & Aydin, T. (2019). Sentiment analysis of Turkish and english twitter feeds using Word2Vec model. 2019 27th Signal Processing and Communications Applications Conference (SIU), 1–4.
• Karimi, S., Metke-Jimenez, A., Kemp, M., & Wang, C. (2015). Cadec: A corpus of adverse drug event annotations. Journal of Biomedical Informatics, 55, 73–81.
• Kumaragurubaran, T., Varshnee, L., Pandi, S. S., & Tharun, K. (2024). Revolutionizing Healthcare Assessment: Deep Learning Strategies to Detect Patient Opinion on Social Media Platforms. 2024 International Conference on Advances in Computing, Communication and Applied Informatics (ACCAI), 1–7.
• Lavanya, P., & Sasikala, E. (2022). Auto capture on drug text detection in social media through NLP from the heterogeneous data. Measurement: Sensors, 24, 100550.
• Lee, J., Yoon, W., Kim, S., Kim, D., Kim, S., So, C. H., & Kang, J. (2020). BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics, 36(4), 1234–1240.
• Li, Y., Tao, W., Li, Z., Sun, Z., Li, F., Fenton, S., Xu, H., & Tao, C. (2024). Artificial intelligence-powered pharmacovigilance: A review of machine and deep learning in clinical text-based adverse drug event detection for benchmark datasets. Journal of Biomedical Informatics, 104621.
• Liu, Z., Lv, X., Liu, K., & Shi, S. (2010). Study on SVM compared with the other text classification methods. 2010 Second International Workshop on Education Technology and Computer Science, 1, 219–222.
• Nishioka, S., Watanabe, T., Asano, M., Yamamoto, T., Kawakami, K., Yada, S., Aramaki, E., Yajima, H., Kizaki, H., & Hori, S. (2022). Identification of hand-foot syndrome from cancer patients’ blog posts: BERT-based deep-learning approach to detect potential adverse drug reaction symptoms. PloS One, 17(5), e0267901.
• Pires, T. (2019). How multilingual is multilingual BERT. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL).
• Rahman, M. M., Shiplu, A. I., Watanobe, Y., & Alam, M. A. (2024). RoBERTa-BiLSTM: A Context-Aware Hybrid Model for Sentiment Analysis. arXiv Preprint arXiv:2406.00367.
• Rawat, A., Wani, M. A., ElAffendi, M., Imran, A. S., Kastrati, Z., & Daudpota, S. M. (2022). Drug adverse event detection using text-based convolutional neural networks (TextCNN) technique. Electronics, 11(20), 3336.
• Saheed, Y. K., & Arowolo, M. O. (2021). Efficient cyber attack detection on the internet of medical things-smart environment based on deep recurrent neural network and machine learning algorithms. IEEE Access, 9, 161546–161554.
• Sharma, A. K., Chaurasia, S., & Srivastava, D. K. (2020). Sentimental short sentence classification by using CNN deep learning model with fine-tuned Word2Vec. Procedia Computer Science, 167, 1139–1147.
• Shen, Z., & Spruit, M. (2021). Automatic extraction of adverse drug reactions from the summary of product characteristics. Applied Sciences, 11(6), 2663.
• Siami-Namini, S., Tavakoli, N., & Namin, A. S. (2019). The performance of LSTM and BiLSTM in forecasting time series. 2019 IEEE International Conference on Big Data (Big Data), 3285–3292.
• SNOMED International. (2024). SNOMED CT – The Global Language of Healthcare. Retrieved from https://www.snomed.org/about-us
• Usha, G., Narang, M., & Kumar, A. (2021). Detection and classification of distributed DoS attacks using machine learning. Computer Networks and Inventive Communication Technologies: Proceedings of Third ICCNCT 2020, 985–1000.