University of Twente Student Theses
As of Friday, 8 August 2025, the current Student Theses repository is no longer available for thesis uploads. A new Student Theses repository will be available starting Friday, 15 August 2025.
How does the integration of syntactic features (POS tags or dependency parsing) during BERT fine-tuning influence the performance of semantic parsing?
Zhang, Jinrui (2025) How does the integration of syntactic features (POS tags or dependency parsing) during BERT fine-tuning influence the performance of semantic parsing?
PDF
1MB |
Abstract: | BERT has achieved outstanding performance in many NLP tasks, but its implicit handling of syntax may limit its effectiveness in shallow semantic parsing. This study explores incorporating explicit syntactic features, such as POS tags and dependency parsing labels into BERT. This is completed through two approaches: addition or concatenation either at the input embedding layer or after transformer layers. Experiments have been conducted, and results show that the additive approach with dependency parsing labels at input embedding layer achieves the highest F1 score, improving performance by 1.24%. This work provides insights into the integration of syntactic features in BERT. Additional Key Words and Phrases: BERT, Semantic Parsing, Large Language Model, BIO tagging, Dependency Parsing, POS tagging. |
Item Type: | Essay (Bachelor) |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 54 computer science |
Programme: | Computer Science BSc (56964) |
Link to this item: | https://purl.utwente.nl/essays/105088 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page