A Study of Low-Resource Medical Named Entity Recognition

This PhD research tackles the problem of limited labeled data in Deep Learning, especially in sensitive fields like healthcare, where expert labeling is costly and privacy-sensitive. The study develops a strategy to improve the identification of key terms (Named Entity Recognition, or NER) from text without needing large amounts of labeled data. It focuses on using prior knowledge, improving the accuracy of the NER model, and enhancing data through augmentation. This approach makes the NER process more effective, even in data-limited medical domains, offering practical and theoretical advancements.