A Model fo Discourse Comprehension : An Integrated Approach of Relevance Theory & Speech Science

在本研究中，我们考虑了如何理解具体语篇中的话语的问题。我们特别关注了日语句子的疑问句形式，它的特点是句末助词“ka”。众所周知，相同的疑问句表达形式可以有不同的解释，这取决于话语的不同语境。正如Itani（1996）所建议的那样，像“Hanako gakuru-ka”这样的疑问句[花子来了吗？]可视为下列问题之一：(i)正题；（ii）考试。问题，（iii）反问句，（iv）说明性问题，(v)猜测问题，（vi）惊喜问题，（vii）提醒，（viii）自我陈述。我们试图从三个层面来解释这些不同的解释，即逻辑语义层面、关联理论层面和声学层面。假设句末助词“ka”的语义不存在语义歧义，在基于高阶逻辑的形式语义框架内对其概念意义进行了分析。很明显，各种上下文信息在对带有“ka”的句子的多种可能的解释中起着重要的作用。我们关注的问题之一是确定在话语中选择特定解释时声学信息的作用。进行了生产试验，测量了各解释层的声学特征。我们发现口译的差异与“ka”音持续时间等声学特征的差异之间存在显著的相关性。结果表明，声学特征可作为判读规范的指标。

In this research we have considered the problem of how to comprehend utterances in a concrete discourse. We have been particularly concerned with the interrogative form of Japanese sentences, which is characterized by a sentence-final particle, "ka". It is well known that a same interrogative form of expression could be interpreted in various different ways, depending on the different contextual setting of discourse. As Itani (1996) suggested, an interrogative sentence like "Hanako gakuru-ka" [Does Hanako come? ] can be taken as one of the following : (i) Genuine question, (ii) Exam.question, (iii) Rhetorical question, (iv) Expository question, (v) Guess question, (vi) Surprise question, (vii) Reminder, and (viii) Self-addressed statement. We have attempted to account for these different interpretations on three levels, i.e., the logical semantics level, the relevance-theoretic level and the acoustic level. Assuming that the meaning of the sentence-final particle, "ka" is not semantically ambiguous, we have analyzed its conceptual meaning within the framework of formal semantics based on higher order logic.It is obvious that various contextual information plays a significant role in many possible interpretations of a sentence with a particle, "ka". One of our concerns was to identify the role of the acoustic information in selecting a particular interpretation in discourse. A production experiment was carried out, and the acoustic features of each interpretation were measured. We found significant correlation between the difference in intepretation and the difference in acoustic features such as the duration of "ka". The results indicate that acoustic features function as an index of specification of interpretations.