Research And Application Of Chinese Discourse Relation Recognition Method

Classroom discourse is the main medium of information interaction between teachers and students in the classroom,and the analysis of classroom discourse is an important way to summarize teaching rules and evaluate teaching quality.As a discourse-granular Chinese style,classroom discourse can be structured with the micro semantics in discourse to present the logic of the classroom discourse and analyze the classroom teaching structure from a micro perspective.Under the current background of the deep integration of artificial intelligence and education,facing inefficiency and high dependency on time and space of traditional methods,how to use artificial intelligence technology to improve traditional classroom discourse analysis methods and provide assistance for teaching research has become an urgent problem to be solved.From the perspective of discourse relation recognition,this paper constructs the discourse structure of classroom discourse by studying the micro-logical and semantic relation between sentences to realizes the intelligent analysis of classroom discourse.Chinese discourse relation recognition is an important research in the field of Natural Language Processing,which aims to study the logical semantic relation between sentences.Discourse relation is divided into explicit and implicit discourse relations according to the presence or absence of conjunction.The implicit discourse lacks clues such as obvious conjunction,so it is necessary to extract the deep semantic features between the argument pairs to construct the logical relation of the discourse arguments,which is more complicated.Therefore,the recognition of implicit discourse relation has become the main task in the current discourse relation recognition.Aiming at the implicit discourse relation,this paper proposes implicit discourse relation recognition methods with enhanced sequence and semantic representation.The discourse structure of classroom discourse is constructed based on the discourse relation.It is applied to the analysis task of classroom discourse,which has achieved good performance.The main work and innovations of thesis include the following three aspects:Firstly,to solve the problem that local sequence information on the granularity of characters and words is easy to be ignored in the existing Chinese implicit discourse methods,this paper proposes a Chinese implicit discourse relation recognition method with enhanced sequence representation.This method first uses Bi-LSTM to model the sentence sequence representation,secondly uses local convolution operation to filter the fine-grained sequence features,and finally uses the semantic code of the attention network as external memory to repeatedly read the important sequence features,and then generates the final argument representation.Experiments on public data sets show that this method exceeded several benchmark models.Secondly,since the implicit discourse has no conjunctive clues,it is necessary to dig deeper semantic information.Therefore,this paper proposes a method to recognize Chinese implicit discourse relation with enhanced semantic representation.This method firstly uses a shallow convolutional network to generate the potential topic information represented by a one-dimensional tensor,and secondly assigns weights to the potential topics and the argument coding of the attention network based on the attention mechanism,and finally generates argument representation with richer semantic information through feature fusion.Experiments on public data sets show that this method exceeded several benchmark models.Finally,this paper constructs a corpus of discourse relations and a discourse analysis system in the field of classroom discourse.In this system,the models based on enhanced sequence and semantic representation are trained in the discourse corpus of classroom discourse.Thereby this system presents the teaching content more intuitively with the structure of the classroom discourse based on discourse relation,which provides a new method and perspective for teaching analysis and evaluation.To solve the problem of Chinese implicit discourse relation recognition,this article proposes effective methods,achieving good results.In addition,the methods are applied in the field of classroom discourse analysis,providing an intelligent auxiliary analysis method for teaching evaluation.