计算机应用与软件2024,Vol.41Issue(2) :257-263.DOI:10.3969/j.issn.1000-386x.2024.02.037

基于关系的函数题目自动解析和解答方法

FUNCTION PROBLEM SOLVING METHOD BASED ON RELATION

孙慧慧 余新国 孟皓 吕小攀
计算机应用与软件2024,Vol.41Issue(2) :257-263.DOI:10.3969/j.issn.1000-386x.2024.02.037

基于关系的函数题目自动解析和解答方法

FUNCTION PROBLEM SOLVING METHOD BASED ON RELATION

孙慧慧 1余新国 1孟皓 1吕小攀1
扫码查看

作者信息

  • 1. 华中师范大学国家数字化学习工程技术研究中心 湖北 武汉 430079
  • 折叠

摘要

关系表示作为数学题目解答的基本问题而广受关注,而对于含有函数的题目解答,由于函数关系本身的复杂性以及表达方式的多样性,使得该问题超出了现有解题范围,为解答提出了新的挑战.针对这一问题,提出一种基于关系的自动解析和解答算法.扩展关系表示方式使其满足函数关系表示与计算的需要;通过改进句法语义模型和提出图形关系模式分别从文字和图形中提取关系;根据函数的模型化定义识别和提取函数关系;利用函数关系与数量关系进行等量代入,消除参数得到解答结果.对采集的数据集进行验证,与基于框架的基线方法相比较,该算法能够获得较好的结果,有效完成了83%的题目理解和66%的正确解答率.

Abstract

As a basic problem of mathematical problem solving,relational representation has attracted much attention.However,for the solving problems with function,because of the complexity of function relationship and the diversity of expression,it is beyond the scope of existing problem-solving and poses new challenges for the domain.Aiming at this problem,this paper proposes a relation-based automatic solving algorithm.The relation representation was extended to meet the needs of function relation representation and calculation.The relations were extracted from the text and graph by improving the syntactic and semantic model and the proposed graph relation pattern.The function relation was identified and extracted according to the model definition of function.The function relation and quantity relation were substituted equally to get the solution result by eliminating extra parameters.Compared with the baseline method based on the solving framework on the collected dataset,the proposed algorithm can achieve better results,effectively completing 83%of the problem understanding and 66%of the correct answer rate.

关键词

数学题目解答/关系解答/函数关系/句法语义模型/自动解答

Key words

Mathematic problems solving/Relation solving/Function relation/Syntactic-semantic model/Automatic solver

引用本文复制引用

基金项目

国家自然科学基金项目(61977029)

出版年

2024
计算机应用与软件
上海市计算技术研究所 上海计算机软件技术开发中心

计算机应用与软件

CSTPCD北大核心
影响因子:0.615
ISSN:1000-386X
参考文献量20
段落导航相关论文