Dowsing for Math Answers

The MathDowsers team's runs at ARQMath

Purpose of the Demo

The task of interest is the Math Community Quesiton Answering (MathCQA) task:

Given a math question, look for potential answers to this math question from an existing math answer database.

This demo displays the task runs for the MathCQA task at ARQMath Lab (Answer Retrieval for Questions on Math) 2020, 2021, and 2022, where the math answer database is the Math StackExchange corpus from year 2010 to 2018, and math questions are selected from the same corpus from year 2019 to 2020.

While the MathDowsers team's runs are displayed by default, interested users might also input custom runs to view their ranked answers. Check the panel to learn more about the features of the demo.

About MathDowsers

MathDowsers is a team of researchers from the University of Waterloo who are interested in dowsing for answers to math questions.

ARQMath-1 & ARMath-2 With the math-aware search engine Tangent-L, the team produces the best participant run of the Answer Retrieval task in the ARQMath Lab in both year 2020 and 2021; and also the best automatic run of the Formula Retrieval task in the Lab in year 2021. The team's code repository can be found here. Full details of the MathDowsers' submissions are described in the following publications:

Yin Ki NG, Dallas J. Fraser, Besat Kassaie, Frank Wm. Tompa. Dowsing for Math Answers, in: Candan K.S. et al. (eds) Experimental IR Meets Multilinguality, Multimodality, and Interaction. CLEF 2021. Lecture Notes in Computer Science, vol 12880. Springer, Cham. https://doi.org/10.1007/978-3-030-85251-1_16

Yin Ki NG, Dallas J. Fraser, Besat Kassaie, Frank Wm. Tompa. Dowsing for Answers to Math Questions: Ongoing Viability of Traditional MathIR , in: CLEF 2021, volume 2936 of CEUR Workshop Proceddings, 2021

Yin Ki NG, Dallas J. Fraser, Besat Kassaie, George Labahn, Mirette S. Marzouk, Frank Wm Tompa, and Kevin Wang. Dowsing for Math Answers with Tangent-L, in: CLEF 2020, volume 2696 of CEUR Workshop Proceedings, 2020

and in the MMath thesis, which includes also follow-up experiments:

ARQMath-3 With a new implemented engine mtextsearch, the team simplifies experimentation, improves retrieval effectiveness, and shows that traditional text retrieval systems can easily be enhanced to become effective math-aware search engines. Full details of the submission are described in the following publication:

Andrew Kane, Yin Ki Ng, Frank Wm. Tompa. Dowsing for Answers to Math Questions: Doing Better with Less, in: CLEF 2022, volume 3180 of CEUR Workshop Proceddings, 2022

More details of the research project can be found in the BrushSearch site.