Seeking Diverse Reasoning Logic: Controlled Equation Expression Generation for Solving Math Word Problems
Seeking Diverse Reasoning Logic: Controlled Equation Expression Generation for Solving Math Word Problems
复制标题
DOI：
10.48550/arxiv.2209.10310
复制
发表时间：
2022-09
期刊：
影响因子：
--
通讯作者：
Yibin Shen;Qianying Liu;Zhuoyuan Mao;Zhen Wan;Fei Cheng;S. Kurohashi
中科院分区：
文献类型：
--
作者：
Yibin Shen;Qianying Liu;Zhuoyuan Mao;Zhen Wan;Fei Cheng;S. Kurohashi
为了解决数学应用题，人类学生利用不同的推理逻辑，得出不同的可能的等式解决方案。然而，主流的顺序到顺序自动解算器的目的是在人工注释的监督下解码固定的解方程。在本文中，我们提出了一种受控方程生成求解器，它利用一组控制代码来引导模型考虑某种推理逻辑，并解码从人类参考转换而来的相应方程表达式。实验结果表明，我们的方法普遍提高了单一未知(Math23K)和多重未知(DRAW1K，HMWP)基准测试的性能，在具有挑战性的多重未知数据集上的准确率提高了13.2%。
To solve Math Word Problems, human students leverage diverse reasoning logic that reaches different possible equation solutions. However, the mainstream sequence-to-sequence approach of automatic solvers aims to decode a fixed solution equation supervised by human annotation. In this paper, we propose a controlled equation generation solver by leveraging a set of control codes to guide the model to consider certain reasoning logic and decode the corresponding equations expressions transformed from the human reference. The empirical results suggest that our method universally improves the performance on single-unknown (Math23K) and multiple-unknown (DRAW1K, HMWP) benchmarks, with substantial improvements up to 13.2% accuracy on the challenging multiple-unknown datasets.