INTERACTION WITH CONTEXT DURING HUMAN SENTENCE PROCESSING
INTERACTION WITH CONTEXT DURING HUMAN SENTENCE PROCESSING
复制标题
DOI：
10.1016/0010-0277(88)90020-0
复制
发表时间：
1988-12-01
期刊：
影响因子：
3.4
通讯作者：
STEEDMAN, M
中科院分区：
文献类型：
--
作者：
ALTMANN, G;STEEDMAN, M
自然语言处理的心理学理论通常假设句子处理器通过选择基于结构标准的单个分析来解决局部句法歧义，例如Frazier（1978）的"最小附件"。根据这些理论，只有在最初的分析后来证明与上下文不一致时，才尝试进行替代分析。(See Ferreira和克利夫顿，1986年;福特、BVresnan和Kaplan，1982年; Rayner、Carlson和Frazier，1983年）。然而，另一种假设是存在的：如果句子被逐渐理解，或多或少是一个词一个词地理解（Marslen-Wilson，1973，1975），那么句法处理原则上可以利用解释是可用的这一事实，使用它们"交互式"地选择amnong替代句法分析，基于它们与上下文的可解释性。本文认为可能的架构，这种增量和互动的句子处理器，并主张一个架构，其中替代分析最初提供并行，然后区分直接呼吁下的选择性或“弱”的相互作用，而不是指令或“强”的相互作用的理解过程。我们注意到，这样的架构并没有以任何方式损害Fodor（1983）的模块化假设。我们回顾实验证据，已声称表明，人类的句子处理是非互动的，介导的纯结构标准。新的结果出现不兼容的结构主义的建议，并支持互动的假设。我们提出了为什么早期相反的结果可能会被打折的原因，并得出结论，人类的句子处理机制解决了修饰语附件歧义诉诸于更高层次的上下文和参考信息的弱相互作用。
Psychological theories of natural language processing have usually assumed that the sentence processor resolves local syntactic ambiguities by selecting a single analysis on the basis of structural criteria such as Frazier''s (1978) "minimal attachment". According to such theories, alternative analyses will only be attempted if the initial analysis subsequently proves inconsistent with the context. (See also Ferreira and Clifton, 1986, Ford, BVresnan, and Kaplan, 1982; Rayner, Carlson, and Frazier, 1983). An alternative hypothesis exists, however: If sentences are understood incrementally, more or less word-by-word (Marslen-Wilson, 1973, 1975), than syntactic processing can in principle exploit the fact that interpretation are available, using them "interactively" to select amnong alternative syntactic analyses on the basis of their plausibility with respect to the context. The present paper considers possible architectures for such incremental and interactive sentence processors, and argues for an architecture in which alternative analyses are initially offerd in parallel, and are then discriminated among by immediate appeal to the comprehension process under a selective or "weak" interaction, as opposed to directive or "strong" interaction. We note that such an architecture does not compromise the modularity hypothesis of Fodor (1983) in any way. We review experimental evidence which has been claimed to show that human sentence processing is non-interactive and mediated by purely structural criteria. New results are presented which appear incompatible with the structuralist proposal, and which support the interactive hypothesis. We suggest reasons why the earlier contrary results may be discounted, and conclude that the human sentence processing mechanism resolves modifier-attachment ambiguities by recourse to higher-level contextual and referential information under the weak interaction.