Relational Networks

语言信息是如何表征的,语言信息是如何在发话和理解中运用的,这两个相关问题是极其重要的语言学理论问题。语言信息的表征不应以能力和运用的剥离为基础,而应以能在语言过程中直接运用的形式为基础,这样的观点更为实际。语言信息一般表征为符号形式,但是如此表征难以表述发话和理解过程。另一种做法是采用关系网络,即采用连通主义模式。在连通主义模式中,语言信息表征于巨型多层网络的连接关系之中。语言的操作就是连通网络关系线和关系符的激活延伸过程。关系网络形式允许我们实际表述各种被忽略的语言信息,该形式还和已知大脑信息一致。

The question of how linguistic information is represented, and the related question of how this information is used in speaking and understanding, are vitally important for linguistic theory. Rather than relying on a separation of competence from performance, as some have proposed, it is more realistic to conceive of the linguistic information as represented in a form that can be directly used in processing. Linguistic information is usually represented in symbolic form, but this practice makes it difficult to account for the processes of speaking and understanding. The alternative is network representation, or connectionism, in which all linguistic information is represented in the connectivity of a large multi-layer network. In this type of system, linguistic processing operates as the traveling of activation among the lines and nodes of the network. The relational network format allows us to account realistically for various kinds of linguistic information that are commonly neglected, and it is also compatible with what is known about the brain.