Leveraging Syntactic Dependency Parsing for Enhanced Semantic Role Labeling in Clinical Text

Semantic Role Labeling (SRL) is crucial for understanding the semantic structure of sentences, particularly in complex domains like clinical text. This paper investigates the efficacy of incorporating syntactic dependency parsing as a pre-processing step to improve SRL performance. We propose a novel architecture that integrates a pre-trained dependency parser with a state-of-the-art SRL model. Our experiments on a benchmark clinical text dataset demonstrate significant improvements in precision and recall compared to baseline SRL models that do not utilize dependency information. Furthermore, we analyze the error patterns of different models and highlight the benefits of dependency parsing in resolving ambiguous semantic roles.