Sentiment Analysis in Greek Clinical Conversations: A Comparative Study of BERT, VADER, and Lexicon Approaches

Sentiment analysis, a subfield of Natural Language Processing (NLP), focuses on the study of emotions expressed in text data. While extensively used in major languages such as English, its application in Greek remains underexplored. We investigate sentiment analysis in Greek clinical dialogues, focusing on hematologic malignancies. These dialogues offer valuable insights into patient experiences, healthcare provider approaches in palliative care, and underscore the critical role of sentiment analysis in accurately assessing patients' emotional status for effective care. Our study compares three methodologies—Greek Lexicon, VADER, and BERT—while addressing the limited Greek resources in healthcare-related NLP. Key findings indicate that BERT outperformed other methods by achieving well-balanced precision and recall in assessing sentiment. The Lexicon-Based approach encountered challenges in identifying both negative and positive sentiments, while VADER showcased robust results. Our study involves data collection, data annotation, model training, and performance evaluation. We aim to address the research gap and the unique challenges of sentiment analysis in Greek, especially in context of hematologic malignancies, where NLP and sentiment analysis resources are limited.