权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Toward the Construction of a Language Model Integrating Syntactic, Semantic and Discourse Information

构建整合句法、语义和话语信息的语言模型

基本信息

批准号：
10610517
负责人：
TAKUBO Yukinori
金额：
$ 2.37万
依托单位：
Kyoto University (2000)Kyushu University (1998-1999)
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
1998
资助国家：
日本
起止时间：
1998 至 2000
项目状态：
已结题

项目摘要

The main purpose of the present research project has been to construct an integrated model of language, incorporating results from syntax, semantics and pragmatics. In order to fully understand the mechanism of language, it is essential that we know how the core computation feeds other cognitive mechanisms, such as concept formation and discourse management. In a language like Japanese, which lacks the major driving force in syntax, i.e. the rich agreement system, most of the interesting things should be found by looking into what happens at the interface level.With those assumptions, we have tried to develop techniques to isolate syntactically licensed phenomena from those that are regulated by cognitive constraints. The structures that are referred to by syntactic licensing are necessarily determined by two factors : merge and linearization. The structural factors must, therefore, be definable by (asymmetric) c-command and/or linear precedence, with concomitant division of labor between the two. If a condition on a construction is outside these two structural factors, it must necessarily be of nonstructural nature. The structural factors are identifiable and judged by native speakers as being categorial, because c-command and linear precedence are categorial notions. The cognitive factors are judged as continuous ; cognitive relations such as proximity, scalar implication etc. are, by its very nature, continuous. We have demonstrated that this is indeed the case by examining the uses of demonstratives, deictic time nouns, conditionals in Japanese and other languages.The results of the present research project were presented at Glow in Asia, the Linguistic Society of Japan, International Conferences on Cognitive Sciences, and other national and international conferences.

本研究项目的主要目的是构建一个综合了句法、语义和语用研究成果的语言模型。为了充分理解语言的机制，我们必须知道核心计算是如何满足其他认知机制的，如概念形成和话语管理。在像日语这样的语言中，缺乏句法上的主要推动力，即丰富的协议系统，大多数有趣的事情应该通过观察界面层面上发生的事情来发现。在这些假设下，我们试图开发出一些技术，将句法许可现象与那些受认知约束控制的现象分开。句法许可所指的结构必然由两个因素决定：合并和线性化。因此，结构性因素必须可以由(不对称的)c指令和/或线性优先来定义，并伴随着两者之间的分工。如果建筑上的条件不在这两个结构因素之外，它必然是非结构性质的。结构因素是可以识别的，并被本族语者判断为范畴，因为c-命令和线性优先是范畴概念。认知因素被判断为连续的；认知关系，如接近、标量蕴涵等，从本质上讲是连续的。我们通过考察指示词、指示时间名词、条件句在日语和其他语言中的用法证明了这一点。本研究项目的结果在Glow in Asia、日本语言学会、国际认知科学会议和其他国内和国际会议上公布。