Examination of cross-modal priming effect in the phoneme level

本研究以日语为研究对象，探讨句法在解决词汇歧义中的作用。特别是，一个后置助词的影响，在两种类型的句子，（a）S-ga-V句（主语，主观后置助词，和歧义动词）和（B）O-wo-V句（对象，客观后置助词，和歧义动词），将被证明的语义访问。本研究中，后置助词的设计是为了消除动词意义上的歧义，如（a）S-ga-Vi。（不及物动词）或（B）O-wo-Vt。（及物动词）。采用跨模态启动方法，即目标名词在视觉上由汉字字符呈现，其出现在一个非正式呈现的词尾动词的词尾前一个音节，或紧接在词尾之后。一个目标名词与主动词的意义和所使用的后置助词都相关（上下文相关），而另一个目标名词只与主动词的意义相关，但与所使用的后置助词无关（上下文无关），第三个目标名词与动词和助词的意义都无关。为了进行这项分析，我们根据受试者语音的数字声谱图对命名潜伏期数据进行了手工编码，因为语音键测量被认为不适合准确测量命名潜伏期。消歧的并行分布式处理（PDP）模型（连接主义模型）非常适合于实现这种基于约束的词汇主义观点，即人类自然语言理解涉及约束满足机制。我们使用PDP模型对该实验进行了计算机模拟。与实验数据一致，PDP模型能够成功地模拟歧义动词的语境依赖和语境独立频率在激活时程中的作用。

This research demonstrates syntactic effects in resolving lexical ambiguity in Japanese, an S-O-V language. In particular, the effect of a postpositional particle on the semantic access of sentence-final ambiguous verbs in two kinds of sentences, (a) S-ga-V sentences (subject, subjective postpositional particle, and ambiguous verb) and (b) O-wo-V sentences (object, objective postpositional particle, and ambiguous verb), will be demonstrated. In this research, The postpositional particles are designed to disambiguate the verbs' meanings as (a) S-ga-Vi. (intransitive verb) or (b) O-wo-Vt. (transitive verb).A cross-modal priming method was used in which a target noun was visually presented by Kanji characters one-syllable before the end, or immediately after the end, of a sentence-final verb presented auditorily. One target noun was related to both the prime verb's meaning and the postpositional particle used (context-dependent), while another target noun was related to only the prime verb's meaning, but it was unrelated to the postpositional particle used (context-independent), and a third target was unrelated to both of the verb's and the particle's meaning. For this analysis, we hand coded the data of naming latencies based on digital spectrograms of subjects' voices, because the voice key measure has been considered unsuitable to measure naming latency accurately.This research has been pursued with reference to the constraint-based lexicalist approach. The parallel distributed processing (PDP) models (connectionist models) of disambiguation are well suited to implement this constraint-based lexicalist view that human natural language understanding involves constraint-satisfaction mechanisms. We performed a computer simulation of this experiment using our PDP model. Consistent with experimental data, the PDP model can successfully simulate the effect of context-dependent and context-independent frequencies of ambiguous verbs in the time-course of activation.