Year: 2025
Type: dissertation
Abstract: In an information-rich environment, automatic summarization helps save time by highlighting key content and enabling faster text navigation. EstSum was enhanced with semantic analysis: two linguistic modules were added, and transfer learning was tested using a few-shot approach to evaluate performance with limited training data. The wordnet-based module slightly improved accuracy but significantly slowed the system, while the language model–based module achieved a negligible increase in accuracy without affecting speed. Although few-shot performance matched the baseline, further improvements are possible by supplying more training examples and refining prompts. To improve accessibility, a web-based user interface was also developed for non-technical users.
Author: Peedosk, Marko
Language: Estonian
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7113227177