Year: 2024
Type: article
Abstract: ABSTRACT In academic literature, the semantic evolution of scientific terms reflects the development of domain knowledge. While there have been studies using diachronic word embeddings to reveal cultural and social shifts, research on the semantic evolution of scientific terms remains limited. This paper proposes a method utilizing diachronic semantic vectors to elucidate the semantic evolution of scientific entities. Firstly, diachronic semantic vectors of entities are trained using incremental learning to compute the semantic shift (ΔA) of a given entity A between periods t1 and t2. Secondly, entities similar to ΔA in periods t1 and t2 are defined as leading reasons and accompanying reasons, respectively. The experiment result shows that the leading reasons can capture the shifts of entities in the nascent stage, and the accompanying reasons can capture the shifts of words in the mature stage. The proposed method offers a better quantitative insight into the details and reasons behind domain knowledge evolution.
DOI: 10.1002/pra2.1134
Authors: Chenxin Du, Chao Ye, Guo Chen
Institution: Nanjing University of Science and Technology
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4403432792