Research On Automatic Analysis And Judgment Of Metaphor Sentences Based On Syntax And Dictionary

Metaphor as a rhetoric is not only a language phenomenon,but also a social phenomenon,it is one of the central language of human language and thinking.Thus the identification and interpretation of metaphor not only involves the literary theory but also involves the thought process and so on many fields.Due to the generality of metaphor coverage and the importance of use,metaphorical automatic processing has become a bottleneck in the field of automatic digest,machine translation,chat robot,information retrieval,computer aided teaching and so on.This article has carried on the thorough and meticulous research around Chinese metaphorical sentence.On the basis of complying with modern metaphor theory,this paper summarizes the four characteristics of metaphorical language and the semantic features of metaphor.The formalization method of metaphor and the classification system of metaphor are given,and a set of metaphorical sentences is proposed and realized.Analysis and determination methods,to large-scale analysis and decision metaphor.On this basis,this paper has developed an automatic analysis and judgment system for large-scale metaphor,and applied for the corresponding protection measures for the system.Experiments show that the ideas and methods proposed in this paper are fruitful.At present,there are many researches on the recognition of metaphorical sentences,such as the use of the role-dependent model to describe and identify the metaphorical sentences in Zhejiang University.Huang Xiaoxi,Zhejiang University,is used in the formal description and classification of metaphorical language.Relations as the basis of formalization,Jia Yu-xiang use of "word forest" and "knowledge network" to identify the sentence sentence.Although the current research has been a lot,but the metaphorical sentence on the candidate body and the candidate of the extraction of the basic problem,either simply rely on dependencies,or manually designated,but the current dependency accuracy is not high,The manual designation avoids the drawbacks of selecting the candidate ontology and the candidate metaphor depending on the dependency relationship,but it does not apply to the application of large-scale metaphor.The research work of this paper is mainly reflected in the following aspects:First of all,this paper puts forward a kind of structure and judgment standard which can understand and formalize the metaphorical rhetoric of the machine.Based on the parting,syntactic analysis and dependency analysis,the structure of metaphor rhetoric is divided into simple and complex forms.This paper,based on the characteristics of metaphor and the structural features of simple metaphor,puts forward the four formal expressions of the simple metaphorical sentence and the four predicate formulas of candidate ontology and candidate metaphor.Secondly,for the complex metaphor sentence,this paper has developed a set of feasible methods to deal with complex metaphor sentences based on syntactic and dependency relations and the axioms of candidate ontology and candidate metaphors based on this formal method,,This paper proposes a method to eliminate the redundant elements based on syntactic analysis and the narrowing method of candidate ontology and candidate metaphor based on dependency relation based on dependency formalization.Finally,after comparing the metaphor judgment method based on"word forest" and"knowledge network" and the merits of the metaphor judgment method based on "knowledge network" and WordNet,this paper chooses the latter as a dictionary of metaphor judgment method and establishes A set of metaphor rhetorical methods of automatic determination and the method of matching the analogue rhetoric of the automatic analysis process.The purpose of this paper is to combine the sentence composition method and the dependency relation,to remove the modified elements of the sentence,to achieve the purpose of obtaining the principal component of the sentence,and to turn the complex metaphor into a simple metaphor sentence to select the candidate body and candidate of the complex metaphor And then use the semantic similarity between the two parts of English and WordNet to calculate the semantic similarity of the two words.Finally,according to the similarity between the lexicon of the metaphorical word,the similarity of the candidate ontology and the candidate metaphor and its characteristics in WordNet,To determine whether the sentence is metaphorical rhetoric and whether it is expressive or metaphorical expression.The method proposed in this paper has got rid of the specific metaphorical sentence and frame dependency,so it has stronger adaptability,and people have great randomness when using metaphor,so this method is also closer to reality.At the same time,this paper first carries out the word segmentation and part of speech annotation,and according to the metaphorical sentence of the characteristics of the annotation of the sentence to deal with the word segmentation and part of speech can not be resolved in the case of further parsing sentence syntax analysis,the syntax analysis can not solve the case The use of dependency relationship to deal with metaphor and then get rid of the dependence on the excessive dependence,thus avoiding the dependency relationship is not high accuracy rate of the limitations of this method.On the other hand,due to the complexity of the Chinese sentence and the diversity of the sentence content and the simultaneous emergence of multiple noun pronouns,many methods have the limitation of sentence length or sentence when recognizing metaphor,Simple or complex metaphor can be judged.After the experiment,the correct rate of this method is 90.13%,the recall rate is 85.53%and the F value reaches 87.76%.The validity and practicability of the algorithm are shown,so we apply to the State Intellectual Property Office The corresponding protective measures.