神经认知语言学“关系网络理论”的多维性探讨  

作　　者：陈豪[1] 梁倩 

机构地区：[1]复旦大学 [2]马德里康普顿斯大学

出　　处：《复旦外国语言文学论丛》2020年第1期1-10,共10页Fudan Forum on Foreign Languages and Literature

摘　　要：本文将以兰姆神经认知语言学"关系网络理论"(RNT)为着眼点,探讨该"网络"构建的神经认知语言体系的多维性,并将其同自然语言处理(NLP)中深度学习基础网络的多维性相比较。本文将从以下几个方面展开讨论:1)关系网络的基本概念;2)关系网络的多维性;3)计算机语言系统多维性探讨(花园幽径句多维性简析);4)深度学习基础网络的多维性在关系网络中的体现等。本文认为,RNT作为具有跨学科属性的研究对象能更好地突破存于神经生物学和计算语言学之间的壁垒,对其多维性的思考将为语言网络观提供一个全新的视角。This article will focus on the "Relational Network Theory"(RNT) in Lamb’s neurocognitive linguistics to explore the multidimensionality of the general neurocognitive linguistic system constructed by the "network", thereto the comparability between the above-mentioned multidimensionality and that of the elementary Deep Learning Networks in NLP. This article will discuss the following aspects: 1) introduction to the basic conception of RNT;2) discussion on the multidimensionality of RNT;3) discussion on the multidimensionality of computer language(brief analysis of a garden path sentence);4) representation of the multidimensionality of Deep Learning Networks in RNT, etc. This article argues that as a research object with its interdisciplinary attributes, RNT can take more advantages to break the barrier between neurobiology and computational linguistics, and its multidimensional reflection will provide a new perspective for the view of language system as networks.

关 键 词：神经认知语言学 自然语言处理 关系网络理论 语言网络观 深度学习网络 

分 类 号：H0-06[语言文字—语言学]