The Research On Natural Language Understanding

Natural Language Understanding is one of the core problems in the field of Artificial Intelligence. From the first Machine Translation system built up in 1954 to now, by the efforts of the scientists of computer science, linguists and psychologists, great achievement have been made in the field of the language understanding for the restricted language and specific range. The achievement have been put into use in many applications, for example, the airline ticket system, weather forecast system etc. However, the goal of the research of Natural Language Understanding, that is, computer can communicate with human being freely, is still a faraway dream.The research of Natural Language Understanding is composed of four parts. From bottom to top, they are language dictionary, syntax analysis, semantics analysis and text analysis. At present, many theories have been proposed in the four parts. For language dictionary there are WordNet of Princeton University, FrameNet of Berkeley University, MindNet of Microsoft, HowNet of Dong Zhendong, Modern Chinese Grammar Dictionary of Yu Shiwen etc. For syntax analysis there are Phrase Structure Grammar and Transform Generation Grammar of Chomsky, Augmented Transition Network of Woods, Tree Adjoining Grammar of Joshi, Dependency Grammar of Tesniere etc. For semantics analysis there are Componential Analysis of Hjelmslev, Case Grammar of Fillmore, Semantics Network of Simmons etc. For text analysis there are Frame of Minsky, Scripts of Schank, Story Grammar of Rumelhart, Rhetoric Structure Theory of Mann etc.There are still a lot of core problems haven't been solved in the research. On one hand, what is natural language in nature and how does it work haven't been revealed yet; On the other hand, the theories of how to build language dictionary and analyze text do not work so well:1 To build a language dictionary to contain as much the knowledge of syntax, semantics and pragmatics for the aim of providing a solid platform for the later work is the bottleneck of the research.2 The theory of text analysis is behindhand compared with the other three parts in the research. Proposed theories seems not to be so effective.This paper proposes new methods to achieve the theories of these two parts. The main work and innovation of this paper are:1 Discusses the history and present status of Natural Language Understanding and the entire framework of Natural Language Understanding including all the four part of language dictionary, syntax analysis, semantic analysis and text analysis and makes a preview of the developing of Natural Language Understanding.2 Using a composed method of cognitive science and artificial intelligence to do the Natural Language Understanding Researching.3 Discusses the principle of how human being cognize the world and issues the difficulties of Natural Language Understanding research from the angel of cognitive science.4 Proposes a new method of language dictionary building base on the cognitive theory.5 Proposes a new method of text analysis using scenes, agents and state chart.6 Develops a prototype system of the new method and analyze it.