Research On Automatic Labeling Algorithm Of Mathematical Knowledge Points Based On CRF And Deep Learning

In the Internet information age,education continues to develop towards the trend of intelligence.Intelligent question bank is an important part of education intelligence,which plays an important role in improving education quality and efficiency.The automatic labeling of accurate test knowledge points is the basis for completing intelligent question bank tasks such as personalized cognitive diagnosis and personalized question recommendation.Therefore,how to accurately and automatically mark the knowledge points of test questions is of great research significance.At present,the research on the automatic marking of knowledge points of test questions mainly focuses on literature and history,while there is still less research on science subjects such as mathematics.Due to the large number of formulas in the mathematics questions,it is difficult to ensure the accuracy of knowledge point labeling by directly using the text classification technology of the general field.Therefore,this paper takes high school mathematics questions as the research object,firstly,according to the particularity of mathematical problems,the extraction and representation of mathematical formulas are studied.Secondly,the classification effect of the TextCNN model,the Transformer model and the TextCNN-Transformer combination model constructed in this paper in the task of automatic labeling of multi-knowledge points in high school mathematics questions is compared and analyzed,and the specific research content is as follows:Firstly,extraction and representation of formulas.This article summarizes the formulas that appear in the test questions into 17 entity categories;After comparative analysis,the Conditional Random Fields(CRF)model is selected to map the formula in the question to the entity category,and then the corresponding entity category is used to replace the formula,so as to remove the noise while retaining the information existing in the original question as much as possible.Secondly,automatic annotation of knowledge points of test questions based on deep learning.This paper integrates the TextCNN model and the Transformer model to construct a TextCNN-Transformer combination model.This model optimizes the feature extraction work by combining the local features of the text extracted by the TextCNN layer and the long-term dependence of the text extracted by the Transformer layer.The experimental results show that in the data set after the replacement of high school mathematical entities,the TextCNN-Transformer combination model is used to automatically label knowledge points,and the accuracy rate and recall rate of 85.48% are achieved.Compared with the TextCNN model and the Transformer model,the evaluation indicators of the model constructed in this paper have different degrees of improvement.The results show that the method constructed in this paper has a good effect on the automatic labeling of knowledge points of test questions,and has certain practical value in students’ personalized cognitive diagnosis and personalized test question recommendation.