Research On Dependency Syntax Analysis Method Based On Neural Network Model In Chinese Compound Sentences

As one of the core topics in the field of Chinese information processing,Chinese dependency parsing has been deeply concerned by researchers.The Chinese compound sentence is also a special sentence type that contains complex structures and semantic information and is widely used,and has a high research value.Therefore,it is very important to study the method of compound sentence dependent syntactic analysis based on neural network model.This thesis mainly focuses on the analysis of compound sentence dependency syntax based on neural network model.The specific research work includes the following three aspects:First,because the dependency tree bank is the basis of statistic-based dependency syntax analysis methods,However,the annotation system of the traditional Chinese compound sentence dependency tree bank lacks in interpreting and dealing with some linguistic phenomena involving parallel structures(joint structures)and complementarity structures.Therefore,based on the revision of the traditional compound sentence dependency tree bank annotation system,this paper constructs a tree bank of Chinese complex sentence dependency.Secondly,for the analysis method based on the neural network model,it is easy to ignore the linguistic structure and result in obvious defects in semantic understanding when analyzing the dependency relationships in clauses of complex sentences.This paper adopts the combination of statistics and rules,and introduces the idea of sentence synthesis analysis in the statistics-based neural network dependency analysis model.The neural network dependency analysis results are matched with the Chinese structure type template for the corresponding rule post-processing,and then the definition of the dependency relationship between the clauses in the complex sentence is combined to finally obtain the result of the dependency analysis of the entire compound sentence.Third,the traditional dependency analysis model based on feedforward neural network has a high accuracy in processing single sentences.However,because the compound sentence is different from the single sentence,its syntactic structure is complex and the semantic expression is rich,which brings certain difficulties to the traditional neural network model in dealing with compound sentence dependency analysis.Therefore,based on the analysis of the influence of collocation constraints and semantic association features on the syntactic parsing of compound sentences,Therefore,this paper improves the traditional model of dependency analysis based on feed-forward neural network by adding new meta-features to feature templates,adjusting neural network parameters,and using cooperative training to train the model,thus further improving the performance of compound sentence dependent syntactic analysis.