Experimen system of Geometry Inference on Web

Web几何推理实验系统

• 批准号: 12680388
• 负责人: HIROKAWA Sachio
• 金额: $ 1.98万
• 依托单位: Kyushu University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2000
• 资助国家: 日本
• 起止时间: 2000 至 2002
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-12680388/
• 关键词: geometry; inference; artificial intelligence; 幾何推論; DCモデル

We have proposed a fast pattern extraction method for geometric inference system. The method plays a core part of our geometric inference engine which runs on Web. A user draws a figure with GUI and specifies the condition and the goal to prove. The ideal figures are obtained as a result of numerical approximation from the roughly drawn figure. Geometric inference are implemented as a production system based on Koedinger's DC-model method which represents a prototypical geometric image. The method needs a precisely drawn figure to detect such patterns. Those patterns is used as guides in the inference process. In this project, we proposed a fast method to detect these patterns in a given figure. We confirmed the efficiency of our method for typical examples of elementary geometry theorems.

我们提出了一种用于几何推理系统的快速模式提取方法。该方法是我们在 Web 上运行的几何推理引擎的核心部分。用户用GUI画图并指定要证明的条件和目标。理想的图形是对粗略绘制的图形进行数值近似的结果。几何推理被实现为基于 Koedinger 的 DC 模型方法的生产系统，该方法表示原型几何图像。该方法需要精确绘制的图形来检测此类模式。这些模式用作推理过程的指南。在这个项目中，我们提出了一种快速方法来检测给定图形中的这些模式。我们证实了我们的方法对于基本几何定理的典型例子的有效性。