Deep learning-based natural language processing techniques for smart healthcare

Abstract

Natural Language Processing (NLP) is one of the most essential technologies for smart healthcare. In recent years, deep learning-based NLP techniques have gathered significant attention. Despite promising results, existing deep learning techniques remain limited due to challenges including the variability and complexity of medical language, the difficulty of integrating external medical knowledge, and the gap between patient and healthcare provider's way of speaking. These challenges lead to issues in healthcare applications. This thesis aims to leverage deep learning-based NLP techniques towards smart healthcare by addressing these challenges. Specifically, a novel classification framework is first proposed to categorize chief complaints from patients' text, leveraging hierarchical clinical department label information to improve classification performance. Second, a medical dialogue generation framework is introduced, modeling patients and doctors separately and integrating external knowledge to generate contextually appropriate patient-doctor conversations. Third, a Rule-Enriched Attention-Based Deep Neural Network is devised to categorize physician responses into distinct social support types, supported by the development of the first dedicated social support lexicon for team-based teleconsultation, improving the quality of online consultations. Finally, a prompt-based is developed to better capture medical entities in clinical text, overcoming challenges posed by complex medical terminology and limited annotated data.

Date of Award	13 Jul 2025
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Zheng LU (Supervisor), Ruibin Bai (Supervisor) & Tieyan Liu (Supervisor)