Research Methods For Dependency Parsing Of Chinese Marked Compound Sentences

Marked compound sentences occupy a momentous position in the Chinese grammar system and have always been a focus of academic attention.The study on them has acquired plenteous achievements in ontology research.However,few of the studies were carried out from the perspective of Chinese information processing.The information processing with complex sentences as the central task has not achieved innovative breakthroughs and progress.The reasons are as follows.First,most studies on compound sentences are qualitative studies from the perspective of human beings,which are human-oriented grammar or semantic analysis.But computers are very difficult to operate.Secondly,the existing studies are based on a certain compound sentence pattern but not on the compound sentence types or the whole compound sentences system.This cannot meet the information processing requirements,which needs the common and unique syntax and semantic rules of all types of compound sentences.To achieve computer-oriented natural language processing,it is necessary to carry out comprehensive and systematic research.In order to adapt to the development of the Chinese information processing and further promote the process of compound sentences information processing,this dissertation is oriented to the Chinese information processing,drawing on the compound sentences theory of Mr.Xing Fuyi,and making a formal description and induction of the marked compound sentences.In this paper,the dependency parsing theory,complexity theory,word embedding Word2vec and other auxiliary means are used to calculate and model the automatic analysis and discriminating problems of the relationship between marker.The analysis of dependency parsing of the marked compound sentences is realized,which lays a foundation for automatically identifying the dependency parsing tree of compound sentences in the future.This paper includes the following four aspects:1.To solve the interference factors of syntax and semantic processing of marked compound sentences.According to our research,these factors include the identification of clause and non-clause paragraph,the relationship between connective markers,syntax structure and pragmatic matching factors,and the pragmatics differences of synonymous compound sentences.This part will discuss and summarize some of these influential factors and given the solutions.2.Construct a basic resource corpus for Chinese information processing of compound sentences.The research of compound sentences needs the support of language knowledge corpus.On the basis of integrating the research results of relevant compound sentences,this paper introduces the theoretical basis of building a knowledge base of the relationship between marker of the marked compound sentences and The work content of database construction,especially the important technical details such as the selection of the relationship between marker,the differentiation of terms and the establishment of features.3.This dissertation introduces the complex network theory and makes a research on the relationship between marker of the Chinese compound sentences and its related research achievements.We analyzes the complex network characteristics of the relationship between marker of the Chinese compound sentences and calculates the average path length,clustering coefficient,and degree distribution of the matching network of the relationship between marker.The research shows that the matching network of the relationship between marker of compound sentences.Complex sentence association mark collocation network has both short(3.625)and high aggregation coefficient C(0.055),which is a kind of small-world attribute.The clustering coefficient of WordNet is smaller than the matching network of the relationship between marker.Then,we should be able to say:the matching network of the relationship between marker.is a small-world properties of complex networks.Its degree distribution does not quite fit the power law distribution curve.So,we think it is not a scale-free network.4.Dependency parsing is used to solve the problems in the semantic analysis of compound sentences.The analytical methods of dependency parsing of compound sentences is to analyze the semantic relationships among various language units in sentence structures and present the semantic relations in the form of a dependency structural tree or dependency schema.The advantage of using semantic dependency to describe the semanteme of a sentence is that there is no need to abstract the vocabulary itself.We only need to describe the vocabulary through the semantic framework of the vocabulary.Paying attention to the semantic relations between the content words in fact or logic,can cross the surface of the sentence changes to the essence of semanteme.Based on the theory of semantic dependency,this paper crosses the constraints of surface syntactic structure of sentences and designs a prototype system of semantic dependency schema of the marked compound sentences.In this paper,the "although...but..." sentence in the transitional complex sentence is used as a individual case to empirically study the dependency structure and semantic relations.The results of the experiment show that our method works well.