FACULTY OF ELECTRICAL ENGINEERING AND COMPUTING
David Dukić · 2025
Read FACULTY OF ELECTRICAL ENGINEERING AND COMPUTING by David Dukić (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-20033.txt.
Introduction
David Dukić wrote this doctoral thesis on adapting pre-trained neural language models for sequence labeling. Sequence labeling gives each token in a text a label. The work focuses on transfer learning, which reuses knowledge from prior training for new tasks. Its test cases include event trigger detection, spotting words that signal an event.
Dukić offers three linked improvements. One adds a signal from a separate, domain free text system inside a shared multi-task model. A second changes the model layout so information can flow both ways across layers. That change targets autoregressive models, which normally read in one direction. A third uses guided learning in context plus response based tuning. Together, the results argue that careful adaptation brings out the best labeling results.
Catalog record
| CREATOR | David Dukić |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 302 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-20033