Question fuzzy-attention embedding Graph-to-Tree network for intelligent math word problem solver
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F26%3AQPSCCV9E" target="_blank" >RIV/00216208:11320/26:QPSCCV9E - isvavai.cz</a>
Výsledek na webu
<a href="http://dx.doi.org/10.1016/j.neucom.2025.130850" target="_blank" >http://dx.doi.org/10.1016/j.neucom.2025.130850</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1016/j.neucom.2025.130850" target="_blank" >10.1016/j.neucom.2025.130850</a>
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Question fuzzy-attention embedding Graph-to-Tree network for intelligent math word problem solver
Popis výsledku v původním jazyce
Designing intelligent solution models for Math Word Problems (MWPs) holds significant importance for intelligent education. The task of solving MWPs involves translating mathematical problem texts into equations. The introduction of artificial intelligence-generated content models has opened new avenues for advancing intelligent education through digital technology, albeit with new challenges. Current intelligent educational assistants often lack adequate mathematical computation and complex reasoning abilities, especially in dealing with specialized domain knowledge. To address this gap, the paper proposes a semantic document embedding method based on axiomatic fuzzy set membership. The method focuses on enhancing the model's mathematical computation capabilities by processing semantic key information embedded in textual problems. The paper introduces an innovative framework for applying intelligent education systems, specifically targeting tasks like quantity relationship graphs in mathematical questions. Extensive experiments were conducted on two popular MWP datasets. Results show that the proposed model significantly outperforms existing baseline models on both benchmark datasets. Further detailed ablation studies demonstrate the effectiveness and interpretability of the proposed model. Moreover, the paper validates the model's reasoning capabilities in practical subjects such as actuarial science and financial analysis. © 2025 Elsevier B.V.
Název v anglickém jazyce
Question fuzzy-attention embedding Graph-to-Tree network for intelligent math word problem solver
Popis výsledku anglicky
Designing intelligent solution models for Math Word Problems (MWPs) holds significant importance for intelligent education. The task of solving MWPs involves translating mathematical problem texts into equations. The introduction of artificial intelligence-generated content models has opened new avenues for advancing intelligent education through digital technology, albeit with new challenges. Current intelligent educational assistants often lack adequate mathematical computation and complex reasoning abilities, especially in dealing with specialized domain knowledge. To address this gap, the paper proposes a semantic document embedding method based on axiomatic fuzzy set membership. The method focuses on enhancing the model's mathematical computation capabilities by processing semantic key information embedded in textual problems. The paper introduces an innovative framework for applying intelligent education systems, specifically targeting tasks like quantity relationship graphs in mathematical questions. Extensive experiments were conducted on two popular MWP datasets. Results show that the proposed model significantly outperforms existing baseline models on both benchmark datasets. Further detailed ablation studies demonstrate the effectiveness and interpretability of the proposed model. Moreover, the paper validates the model's reasoning capabilities in practical subjects such as actuarial science and financial analysis. © 2025 Elsevier B.V.
Klasifikace
Druh
J<sub>SC</sub> - Článek v periodiku v databázi SCOPUS
CEP obor
—
OECD FORD obor
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Návaznosti výsledku
Projekt
—
Návaznosti
—
Ostatní
Rok uplatnění
2025
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Údaje specifické pro druh výsledku
Název periodika
Neurocomputing
ISSN
0925-2312
e-ISSN
—
Svazek periodika
650
Číslo periodika v rámci svazku
2025
Stát vydavatele periodika
US - Spojené státy americké
Počet stran výsledku
31
Strana od-do
130850
Kód UT WoS článku
—
EID výsledku v databázi Scopus
2-s2.0-105009776901