| This dissertation is a part of the systematic project oriented to Natural Language Processing, the Construction of the Chinese FrameNet based on Frame semantics (by Fillmore) and supported by corpus evidence. It aims at (1) deciding preliminarily the framework and methods for building the CFN, based on the introduction of some linguistic theory and other similar work and (2) carrying out the methods in a specific research on building a small CFN database in a narrow domain. There are 6 chapters.Chapter 1 discusses, in a broad outline, the trend of Natural Language Processing on the background of Cognitive Science in terms of the current level of the study. Based on it, the construction of a semantic knowledge database for Chinese is chosen as the subject of this dissertation. After a survey of several semantic knowledge database projects, the principles of Frame Semantics theory and the Berkeley FrameNet database are confirmed as the best lexical resources for NLP research with the development of Cognitive Science. Then the research objectives are defined as using Frame Semantics theory and introducing the principles and system of the FrameNet to solve the problems in the construction of the CFN, according to the features of Chinese language.Chaper 2 introduces the basic theory of Fillmore's Frame Semantics in terms of its usefulness for the CFN applications and uses the FrameNet results to direct the construction of the CFN, after a survey of its semantic representations and its specific data.In terms of the whole research on the CFN construction, Chapter 3 shows the preparation for the corpus and tools, chooses proper approaches for the construction after some comparison and deduction and makes the basic principles for it as an extension of the methodology borrowed from the FrameNet project.Using the methodology and approaches above, Chapter 4 puts effort into constructing Chinese frame database for Chinese cognition predicates. First, it defines the research objectives by categorizing cognition predicates and then demonstrates, using examples, how to classify and define frames and how to describe the frame semantic information for them. Chapter 5 generalizes the methods for annotating real sentences with semantic roles, using the descriptions in the cognition Frame database and the methods for constructing the Lexical database. For sentences annotation, three aspects are investigated, grammatical function, phrase type and frame element, with emphasis focusing on the conflict between syntax and semantic structure. In addition, it compares the annotation results with other semantic annotated corpus to demonstrate the advantages of our work. On the Lexical database, the manners of describing the meanings of lexical items and the formalization of annotation report are discussed. The results of the lexical database show its significance for NLP applications.Chapter 6 draws conclusions of the research, including the specific achievements and some general ideas about the linguistic research. Some subjects of futher research is also proposed. On the theoretical aspect, this dissertation introduces, for the first time, FiUmore's Frame Semantics into Chinese study. On the practical, it uses corpus method to construct the CFN, paving a way for the whole research project with specific results. The methods of CFN construction and the specific data can be used directly for various Chinese Information Processing applications. For the linguistic research, it provides a rich description of semantics, deepening the recognition of the language. |
PDF Full Text Request