Research And Implementation Of Chinese Complex Sentence Judgment And Semantic Relation Recognition System

Chinese complex sentences play an important role in natural language.They are connected with chapters and contain clauses.They also have syntactic and pragmatic properties.The recognition of semantic relations is the key to semantic analysis of complex sentences.Complex sentences can be divided into explicit and implicit complex sentences according to the presence or absence of Related words.Chinese complex sentences’ automatic judgment and recognition of semantic relations contribute to the development of Chinese abstract semantic representation,information extraction,machine translation,automatic question answering and other fields.Due to the rich semantic relations between the clauses of complex sentences,it is difficult to recognize them automatically,and the lack of connectives makes the task of judgment and recognition more difficult.In this thesis,the deep learning method is used to explore the two tasks of Chinese complex sentence judgment and semantic relationship automatic recognition,and then the joint model method is used to improve the recognition performance.Finally,in order to intuitively show the results,the system of Chinese complex sentence and semantic relationship recognition is realized combined with engineering practice.The main work is as follows.(1)Research on the judgment of complex sentences and the recognition of semantic relations of complex sentences.For the coding part of the two studies,Bi-LSTM is used to determine complex sentences,while Tree-LSTM is used to identify semantic relations.In complex sentence decision,Bi-LSTM is used to capture the context semantic information of the sentence,and attention mechanism is used to mine deeper semantic information.Then CNN is introduced to extract the local information of the sentence,and softmax is used to predict the probability of complex sentence.The first step of semantic relation recognition is to use Stanford parser to analyze syntax to get syntax tree,add tag information,and use Bert to model semantics to get rich semantic information.Then,Tree-LSTM is used to encode.In addition to capturing semantic information,sentence structure information can also be used.Finally,softmax is used to predict.The experimental results show that the two methods can effectively identify Chinese complex sentences and semantic relations.(2)A combined model of complex sentence judgment and semantic relation recognition based on combined neural network is proposed.In order to ensure the similarity of the parameters of the two tasks,a joint model is formed by combining the judgment of complex sentences with the recognition of semantic relations.The input processing uses the above syntactic analysis and the best model to get the word vector,and then uses Bi-LSTM and Tree-LSTM to encode the word vector,and uses the attention mechanism and sentence matching to enhance the relevance between sentences and obtain rich semantic information,so as to improve the performance of the model.Experiments show that the joint model with sentence matching and attention mechanism can achieve good results.(3)Design and implement the system of Chinese complex sentence judgment and semantic relation recognition.Based on Flask framework,Docker,HTML,JS and other technologies.The system can be divided into three parts: model deployment,complex sentence judgment and visualization.The main function is using the deployed model to identify the input,then present the results and syntax analysis diagram in the system interface.