Research On Application Of Improved Topic Segmentation Model In Teachers' Discourse Text Analysis

As an important medium and main method of classroom teaching,classroom discourse is also one of the important indicators to judge teachers' professional quality,teaching philosophy and teaching effect.The structured and systematic analysis of classroom teaching discourse content using schematized and quantified methods can scientifically and objectively present the inner level logic of classroom teaching and analyze the structure of classroom teaching from a micro perspective.At present,most teaching discourse analysis adopts time-consuming and labor-intensive methods such as manual observation,interviews,and questionnaires.How to improve traditional teaching analysis methods and assist educators to efficiently grasp and analyze the characteristics of teachers' teaching is an urgent problem.Aiming at the current lack of intelligent analysis methods for Chinese education texts in this field,this paper proposes a teacher discourse analysis method based on an improved topic segmentation model.This model is used to segment the topic of teacher discourse texts and perform subtopics.Paragraphs are analyzed for content structure and visualized.The experimental results show that this method is more effective than other traditional methods in the topic segmentation tasks of discourse texts.It also realizes the intelligent processing of topic segmentation and structural analysis of classroom discourse texts and can be effectively applied to the analysis tasks of classroom teaching texts.In this paper,the technical research for classroom discourse text analysis mainly includes the following three aspects:First of all,in view of the problem that the traditional LDA topic model is not accurate enough to express word relevance,this paper proposes to use word2vec word embedding model to obtain word context information,fuse global topic distribution features and local context semantic features,and further clarify the implied semantic relationship between corpus.Using the obtained topic distribution and semantic information,similarity measures are performed on each text unit in the document,and the best topic segmentation boundary is selected by the local minimum boundary recognition strategy to achieve the semantic paragraph division of the text.The experimental results on the discourse text in the Chinese classroom show that the method has better results than other unsupervised topic segmentation models.Secondly,in order to further explore the topic structure relationship of teachers'discourse content under each sub-topic,this paper proposes to use the technology of visualizing the topic structure relationship based on the lag sequence analysis method.This method uses the TFIDF word weight algorithm to extract the subject words from the document,calculates the position distribution sequence of the subject words,and performs a lag sequence analysis on the subject words through the feature sequence to obtain the transformation probability matrix.The visualization technology is used to transform the mathematical information into logical clarity.The structure diagram of teaching content provides a new perspective for educators to analyze teaching quality.Finally,in order to effectively verify and apply the teacher's discourse analysis method proposed in this paper,this article also builds a corpus for text segmentation of teaching discourse.The real course videos were manually transcribed and annotated,and the data were selected and partially annotated by using baidu baike and the corpus of high-quality online lecture scripts,so as to provide data support for the specific practice and test of teacher discourse analysis.