论文部分内容阅读
本文研究了支持规则推理的神经网络模型，表明通常执行的推理与符号系统在方法上确实相似，只是它们对常识推理提供了更多的方法。CONSYDERR是一种支持常识推理的连接结构，其目的是给出常识推理的一种模型，并纠正传统规则系统中的脆弱性问题。本项工作表明，推理的连接模型不仅实现了符号推理，而且是一种更好的常识推理的计算模型。
This paper studies the neural network model that supports rule-based inference, and shows that the commonly performed inference and symbolic systems are methodologically similar, but that they provide more ways for common-sense reasoning. CONSYDERR is a join structure that supports common-sense reasoning, and its purpose is to give a model of common-sense reasoning and to correct the problem of vulnerability in traditional rule-based systems. This work shows that inferential connectivity models not only enable symbolic inference, but also a better model of common-sense reasoning.