系统功能语法的形式化研究

发布时间：2018-07-02 12:21

本文选题：系统功能语法 + 形式化　；参考：《江南大学》2014年硕士论文

【摘要】：系统功能语法是一种面向应用的语言学理论，韩礼德称其为“适用语言学”，其理论已经广泛地应用于各种不同的领域，例如文体学、翻译、外语教学等等。在各种应用领域中，韩礼德非常重视人工智能，而当系统功能语法应用于人工智能时，所面临的问题就是对语法理论进行形式化研究。他认为能否将语言理论应用于人工智能是检验其正确性的重要手段，同时也是使之发挥更大作用的好机会。至今为止，在国外系统功能语言学的形式化研究中已经提出了三种主要的方法：1)二十世纪五六十年代，韩礼德首次将其创造的阶与范畴语法用于机器翻译、检索以及词典编纂等。2）七十年代，Winograd第一次基于系统语法建立了一个自然语言处理系统——SHRDULE。至此，许许多多的自然语言生成系统应运而生，包括PROTEUS, PENMAN, SLANG, COMMUNAL等。3）八十年代以后，系统功能语法的应用转向语料库，包括口语语料库和多模态语料库。在此期间，韩礼德探索了语义计算与人工智能结合的可能性。在这些形式化研究中，Matthiessen等人的工作具有较大的影响力，受到了语言学界和计算机界的共同关注。本文将研究系统功能语法及其改进途径，结合Penman语篇生成系统，重点研究Matthiessen等人提出的Nigel语法并改进其中的系统网络表述法。主要内容包括1）Penman系统的工作流程；2）Nigel语法的分层及其与环境的交互；3）系统网络形式化方法及其改进途径。在研究的过程中主要采用了两种研究方法。第一种为文献法，，主要搜索了许多有关系统功能语法在自然语言处理系统中的文献资料，另外与计算语言学和人工智能相关的资料也略有收集。第二种为形式化的方法，由于现实的自然语言极为复杂，不可能作为计算机的处理对象，为了使现实的自然语言成为可以由计算机直接处理的对象，在这众多的应用领域当中，我们都需要根据处理的需求，把自然语言处理抽象为一个“问题”，再把这个问题在数学上加以“形式化”，建立语言的“形式模型”，使之能以一定的数学形式，严密而规整地表示出来以至在计算机上实现。通过这项研究，本文发现：1、在传统的系统功能语言学理论研究中，学者们一般偏重于功能语言学；但是在面向自然语言处理用途的研究中，则必须注重系统语法，因为系统语法本质上就是该语言学理论的一种形式化方法。2、可以更好地理解系统功能语言学与语料库语言学结合的动因，它可以更有效地弥补传统的系统网络研究中的不足，将一个可能性系统发展成为概率系统。然而，工作中还存在一些局限和不足，比如现在语言学研究重点已经转向多模态，而此项研究目前还处于文本阶段，需要长足的进步发展；或者是否可以考虑运用系统功能语法与汉语结合进行考察，因此未来的阶段我们可以考虑转向计算机实现和与汉语结合的研究方向。
[Abstract]:Systemic functional grammar is an applied linguistic theory, which Halliday calls it "applicable linguistics". It has been widely used in various fields, such as stylistics, translation, foreign language teaching and so on. Halliday attaches great importance to artificial intelligence in all kinds of application fields, but when the system functional grammar is applied to artificial intelligence, the problem is to make formal research on grammar theory. He believes that whether language theory can be applied to artificial intelligence is an important means to test its correctness, but also a good opportunity to make it play a greater role. Up to now, three main methods have been proposed in the formal studies of systematical functional linguistics abroad: (1) in the 1950s and 1960s, Halliday first applied the order and category grammar created by Halliday to machine translation. In the 1970s, Winograd first built a natural language processing system-SHRDULE based on system grammar. So far, many natural language generation systems have emerged, including proteus, Penman, Slange, COMMUNAL, etc.) since 1980s, the application of system functional grammar has turned to corpus, including spoken corpus and multimodal corpus. During this period, Halliday explored the possibility of combining semantic computing with artificial intelligence. In these formal studies, Matthiessen's work has a great influence, and has attracted the common attention of the linguistic and computer circles. In this paper, we will study the functional grammar of the system and its improvement approach, combining with the Penman text generation system, focus on the Nigel grammar proposed by Matthiessen and improve the representation of the system network. The main contents are as follows: 1) the workflow of Penman system 2) the delamination of Nigel syntax and its interaction with the environment 3) the formal method of system network and its improvement approach. In the process of research, two research methods are mainly used. The first is literature method, which mainly searches for a lot of documents about system functional grammar in natural language processing system, and also collects a little data related to computational linguistics and artificial intelligence. The second method is formalized. Because the real natural language is very complex, it can not be used as the processing object of the computer. In order to make the real natural language become the object that can be processed directly by the computer, it is in many application fields. We all need to abstract natural language processing into a "problem" according to the needs of processing, and then "formalize" the problem in mathematics and establish a "formal model" of the language so that it can be formulated in a certain mathematical form. It is expressed in a strict and regular manner and can be implemented on a computer. Through this study, it is found that in the traditional theoretical study of systemic functional linguistics, scholars tend to focus on functional linguistics, but in the study of natural language processing, we must pay attention to system grammar. Because systemic grammar is essentially a formalized method of this linguistic theory, it can better understand the motivation of the combination of systemic functional linguistics and corpus linguistics, and it can make up for the shortcomings of traditional systematic network research more effectively. A probabilistic system is developed into a probabilistic system. However, there are still some limitations and shortcomings in the work. For example, the focus of linguistics research has turned to multi-modal, and this research is still in the textual stage, which needs considerable progress. Or whether we can consider the use of system functional grammar and the combination of Chinese for investigation, so in the future, we can consider turning to the research direction of computer implementation and combination with Chinese.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：H04

【参考文献】