融合文本与图形理解的电路题目自动解答

AUTOMATIC SOLVING OF CIRCUIT PROBLEMS BY FUSING TEXT AND SCHEMATIC UNDERSTANDING

电路题目自动解答是人工智能领域前沿研究问题。提出一种融合文本和图形抽取物理关系的电路题目自动解答新方法。通过句法语义模型抽取题目文本中的数量关系,再使用网孔搜索算法抽取电路图形中的结构关系,从而形成一致性题目理解。为了验证该方法的有效性,在电路题目数据集上分别设计了文本、图形的理解及自动解答对比实验。结果表明:句法语义模型对电路文本关系完全抽取率达97.22%,电路图形中的VCR、KCL和KVL关系抽取准确率分别为90.91%、81.82%、91.3%,而文本和图形融合实现的电路题目自动解答,解答率达88.89%,验证了该方法的有效性。

Automatic solving of circuit problem is a frontier research problem in the field of AI.This paper proposes a new method of automatic solving of circuit problem by fusing textual and graphical extraction physical relations.The method extracted the quantitative relations in the subject and text by the syntax-semantics model,and then used the mesh search algorithm to extract the structural relationship in the circuit schematic,so as to form a consistent problem understanding.In order to validate the effect of our method,a series of comparative experiments were designed on the basis of circuit problem data set,including text understanding,schematic understanding and automatic solving.The full extraction rate of the syntax-semantics model for circuit text relations was 97.22%,and the extraction accuracy of VCR,KCL and KVL relations in the circuit graph was respectively 90.91%,81.82%,91.3%.While the circuit problem oftext and schematic fusion was automatically solved,and the answer rate was 88.89%.The experimental results prove the effectiveness of the new solution.