Year: 2025
Type: conference-paper
Abstract: In mathematical literature, terms can have multiple meanings based on context.Manual term disambiguation across scholarly articles demands massive efforts from mathematicians.This paper addresses the challenge of automatically determining whether two or more definitions of a mathematical term are semantically different.Specifically, the difficulties of understanding how contextualized textual representation can help solve the problem are investigated.A new dataset MathD2 for mathematical term disambiguation is constructed with ProofWiki's disambiguation pages.Then three approaches based on contextualized textual representation are studied: (1) supervised classification based on the embeddings of concatenated definitions and titles; (2) zero-shot prediction based on semantic textual similarity (STS) between definition and title and (3) zeroshot LLM prompting.The first two approaches achieve accuracy greater than 0.9 on the ground truth dataset, demonstrating the effectiveness of our methods for automatic disambiguation of mathematical definitions.Our dataset and source code are available here: https://github. com/sufianj/MathTermDisambiguation.
Authors: Shufan Jiang, Mary Ann Tan, Harald Sack
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412944219