ABSTRACT
In automatic solution for mathematical problem by machines, problem understanding is an indispensable prerequisite to answer correctly. In the current methods, problem understanding refers to obtaining equivalent representations of the problems, such as linear equations, polynomials, and so on. However, in the solution process, the problem equivalent representation does not guarantee semantic interoperability with the problem, and it is not easy to generate a human-like solution by backtracking the solution process. Therefore, creating a clear equivalent representation of the question is the direction of the automatic solution. Based on the WordEmbedding technology, this paper proposes a new method to convert the math problems of natural language descriptions into equivalent semantic vector representations that are easy to understand by computers, and to identify the types of problems based on deep learning. In this paper, we take the famous Chinese ancient math problem of "chicken and rabbit in the same cage" as an example, and obtain a good recognition rate of 72%.
- Alan Mathison Turing et al. Computing machinery and intelligence.Google Scholar
- D.G. Bobrow, Natural language input for a computer problem solving system [M]. Report MAC-TR-l, Project MAC, MIT, Cambridge, 1964.Google Scholar
- 吴文俊,初等几何判定问题与机器化证明 [J]. 中国科学 (A), 1977: 6: 507--516 (Wu Wenjun, elementary geometric decision problems and proof of mechanization [J]. Chinese Science (A), 1977:6:507--516)Google Scholar
- 1982 年,张景中 ≪面积关系帮你解题≫ (In 1982, Zhang Jingzhong ≪area relationship helps you solve problems)Google Scholar
- J.Z. Zhang, S.C. Chou, X.S.Gao. Automated production of traditional proofs for theorems in Euclidean geometry. The Hilbert intersection point theorems. Dept. of Computer Science, WSU, TechReP:TR-92-3, 1992.Google Scholar
- W.Kinsch, J.G. Greeno. Understanding and solving word arithmetic problems, Psychology Review, 1985.Google Scholar
- 马玉慧,基于认知模型的数学应用题自动解答研究,中央 编译出版社, 2012年4月. (Ma Yuhui, research on automatic solution of mathematical application problems based on cognitive model, Central Compilation&Translation Press, April 2012)Google Scholar
- N. Kushman, Y. Artzi, L. Zettlemoyer and R. Barzilay. Learning to automatically solve algebra word problems. Association for Computational Linguistics(ACL), 2014Google ScholarCross Ref
- 吴文俊,几何定理机器证明的基本原理:初等几何部分 [M]. 科学出版社, 1982. (Wu Wenjun, basic principle of machine proof of geometric theorem: elementary geometry [M]. Science Press, 1982)Google Scholar
Index Terms
- A Novel Approach for Understanding Primary Mathematical Problems Described in Natural Language
Recommendations
The naproche project controlled natural language proof checking of mathematical texts
CNL'09: Proceedings of the 2009 conference on Controlled natural languageThis paper discusses the semi-formal language of mathematics and presents the Naproche CNL, a controlled natural language for mathematical authoring. Proof Representation Structures, an adaptation of Discourse Representation Structures, are used to ...
Menu-based natural language understanding
AFIPS '84: Proceedings of the July 9-12, 1984, national computer conference and expositionMenu-Based Natural Language Understanding is a new approach to building natural language interfaces. It retains the main goals of natural language systems: flexibility, expressive power, learnability and mnemonicity. However, it solves most of the ...
An approach to program understanding by natural language understanding
An automated tool to assist in the understanding of legacy code components can be useful both in the areas of software reuse and software maintenance. Most previous work in this area has concentrated on functionally-oriented code. Whereas object-...
Comments