Medizinische Universität Graz Austria/Österreich - Forschungsportal - Medical University of Graz
Gewählte Publikation:
Kugic, A.
Leveraging Machine Learning Methods for Processing Non-Lexical Content in Clinical Narratives
Doktoratsstudium der Medizinischen Wissenschaft; Humanmedizin; [ Dissertation ] Medizinische Universität Graz; 2025. pp. 117
FullText
- Autor*innen der Med Uni Graz:
- Betreuer*innen:
- Kreuzthaler Markus Eduard
- Schulz Stefan
- Altmetrics:
- Abstract:
- INTRODUCTION. Any kind of narrative in electronic health records, regardless of the language, contains specialized forms of expressions. These often include elements from multiple languages, particularly Latin, Greek and English. The use of jargon expressions, short forms, and other linguistic features require specialized processing pipelines to extract and disambiguate clinical mentions. The main challenge to overcome is that these features are often non-lexical and fully context-dependent. METHODS. This dissertation used various datasets and computational frameworks for machine learning investigations, such as clinical terminologies, domain dictionaries, short form sense inventories, language models and natural language processing frameworks. Each investigation underwent rigorous testing including annotation of results by domain experts to correctly evaluate the performance of each subtask.