基于语义的词汇判断的计算模型

A COMPUTATIONAL LEXICAL DECISION MODEL BASED ON SEMANTICS

词汇判断是一种在字词认知研究中广泛使用的实验范式。本文提出了一个基于分布式语义表征的计算模型，用来解释某些在词汇判断作业中出现的实验现象。模型是一个包括词形层、隐层、语义层、词典层和判断层的前传网络。词形层、隐层和语义层之间的连接权重在学习过程中被不断调整，使用反传算法进行学习。语义层、词典层和判断层之间的连接权重在学习前就已被固定，在学习过程中不进行调整。模型模拟了在词汇判断作业中出现的四种实验现象：（１）频率效应；（２）语义启动效应；（３）频率与语境的交互作用；（４）重复启动效应。模型所表现出的行为特性由模型的框架和学习算法所决定。

Lexical decision is a widely used paradigm in research on word recognition.A computation model based on distributed semantic representation was presented to explain the major phenomena which were found in lexical decision task.The model consisted of a forward network of five sets of orthographic,hidden,semantic,lexicon and decision units.Weights on connections between orthographic, hidden and semantic units were modified during a training phase using the back propagation learning algorithm.Weights on connections between semantic,lexicon and decision units were fixed before the learning process.The model simulated some aspects of human performance in lexical decision task,including a)frequency effect;b)semantic priming effect;c)the interaction between frequency and context;d)word repetition effect;e)degradation effect.The performance of the model came from the architecture of the model and the learning rule.