常识知识的思维机理-《计算机技术与发展》

文章信息/Info

Keywords:: humanoid resolving; commonsense knowledge; thinking mechanism; semantic identification; logic calculation

摘要:: 人类思维是以大量常识知识记忆和浅层有限推理为特征的,因此,类人思维机器实现应该着力发展知识存储和思维方式的研究。该文对于常识知识的处理过程包括语义识别和逻辑计算两个部分,语义识别又分为常识知识表示和匹配检索,而逻辑计算又包含以规则推理为核心的基本逻辑计算、以场景模式类比为核心的类比逻辑计算和以试探枚举为核心的创造性思维计算。首先讨论了常识知识的思维模型,然后以小学数学应用题类人机器求解为例,详细讨论了以数据配置为基础的语义识别方法,解释了类比逻辑计算和创造性思维计算的机器实现过程。最后,总结了后续技术要点,并提出了提高常识知识表示和积累效率的工程研究方向。

Abstract:: Human thinking is characterized by massive commonsense knowledge memory and shallow limited reasoning. Therefore,the machine realization of humanoid thinking should focus on the development of commonsense knowledge storage and thinking mode. The processing workflow of commonsense knowledge includes two parts:semantic identification and logic calculation. Semantic identificationis divided into commonsense knowledge representation and matching retrieval. Logic calculation includes basic logic calculation with rule reasoning as the core,analogy logic calculation with scene pattern analogy as the core,and creative thinking calculation with exploratory enumeration as the core. The thinking model of commonsense knowledge is discussed firstly. Then,taking the humanoid resolving of primary school mathematics application problems as example,the semantic recognition method based on configuration files is discussed indetail,and the machine implementation workflow of analogy logic calculation and creative thinking calculation is explained. Finally,the following technical points are summarized,and the engineering research direction to improve the efficiency of knowledge representation and accumulation is proposed.