Research On The Classification Method Of Textbook Moral Items Based On Deep Learning

In recent years,with the rapid development of Internet technology based on big data,major Internet platforms are generating a large amount of data information all the time,and text data information is one of them.How to extract the hidden value from these complex text data Information is a research hotspot in many industries today.Text classification extracts the content feature information of a piece of text through a model and automatically classifies it into a set category,which provides a strong support for computer processing of text data.Text classification can also be applied to the moral education classification method of textbooks researching.Moral education is the cornerstone of personal development.My country's compulsory education law emphasizes that schools should put moral education in the first place in education and teaching.Textbooks are an important medium for students to carry out moral education.However,at present,target of moral education relies on manual work,which has the problem of strong subjectivity and low efficiency.In response to this problem,this paper proposes an efficient short text classification model for teaching materials,IoMET_A(Indicators of Moral Education Target based Attention,IoMET_A),use IoMET_A to perform deep learning on the short text dataset of Shanghai textbooks,realize automatic classification of textbook moral indicators,and evaluate model performance through F1-measure.Experiments show that IoMET_A can better complete the task of classifying the moral indicators of textbooks.The specific work of this paper is as follows:The text preprocessing method is studied.Compared with English text classification,Chinese text preprocessing is a very critical part in the process of building a classification model,including a series of tasks such as data cleaning,word segmentation,stop word removal,and data enhancement.In this paper,a new data enhancement technique is proposed by combining SMOTE,EDA and other algorithms,and a data set with more balanced sample distribution and higher text quality is obtained,so that the model has better generalization ability.Word vector representation and feature extraction methods are studied.Text is unstructured data information,so how to quantify text and feature extraction has become one of the most basic research tasks in the field of natural language processing.For this reason,this paper studies One-hot,word2vec,and Glove in the short text classification task of textbooks,discover IoMET_A and word2vec has a significant effect.In order to make better use of the connection and location information between words,the IoMET_A model is established by combining the Attention mechanism and TextCNN,and the advantages and disadvantages of the two complement each other,so that the model can more fully extract the feature information of the text,and a series of comparative experiments have verified that the IoMET_A classification effect is better.