Answer Retrieval for Math Questions Using Structural and Dense Retrieval

Download PDF
Open Webpage

Wei Zhong, Yuqing Xie, Jimmy Lin

Published: 2023, Last Modified: 08 Jan 2026CLEF 2023EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Answer retrieval for math questions is a challenging task due to the complex and structured nature of mathematical expressions. In this paper, we combine a structure retriever and a domain-adapted ColBERT retriever to improve the effectiveness of math answer and formula retrieval. We find these two approaches generate highly effective outcomes because structure search can use unsupervised structure similarity as a strong prior signal to math document relevance, and the ColBERT retriever is able to capture contextual similarity and semantic matching effectively by finding additional relevant math contents even if they are using different formulas.