Year: 2024
Type: conference-paper
Abstract: Etymology, and the field of lexicography, is often constrained by unstructured data formats buried in scholarly articles and dictionaries.This paper presents a methodology and an empirical study for creating a structured etymological dataset suitable for computational analysis.Using data from the Online Etymology Dictionary (Etymonline), we manually annotated a subset of entries to establish a high-quality ground-truth dataset and fine-tuned the FLAN-T5-base model to extract structured etymological relationships automatically.The resulting dataset contains over 103,000 relationships covering 63,603 English lexical terms.Our findings emphasise feasibility of using large language models for structuring lexicographical data, exploring the transferability of the model to other dictionary datasets with no additional manual annotation.
Authors: Yuan Gao, Weiwei Sun
Institution: University of Cambridge
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Funder: University of Cambridge
OpenAlex ID: W4402670486