【摘要】 分别分析了传统的语义网络和人工神经网络方法在知识表示方面的特点和不足,提出了将两者结合起来构建具有语义单元和神经单元双重机能的语义神经单元的设想。以此为基础,构造出具有全连通结构的语义神经网络,给出了网络的权值学习方法及概念单元的语义联想机制,从而形成自主学习与语义联想相统一的集成化知识表示结构。它既能对概念语义及其关联关系进行直观、准确的表示,同时又对概念语义的联想、学习和更新等过程提供统一的支持平台。更多还原

【Abstract】 This paper suggestes a new structure to represent conceptual knowledge,named neural-semantic cell,which combines the explicitness of semantic network with the dynamic of artificial neural network together.We assume the cell consisting of semantic parts and status parts.The former is responsible for declaring and explaining the meaning of the concept,and the latter for computing the relationship between difference concepts and controlling the status of the cell. Moreover,based on the neural-semantic cell,a neural-semantic network springs from the spreading activation model.The neural-semantic network is constructed as a connected graph and is embedded the mechanism to learn from outer independently and the function of semantic association among internal cells.Thus the network cannot only represent explicitly the semantic information of the concepts,and also image dynamically the relationship between concepts.By the leaning mechanism,the neural-semantic network can enrich its conceptual knowledge and evolve the semantic information of the conceptual cell.更多还原