| In recent years,artificial intelligence technology has continued to develop in the field of education.More and more schools and training institutions have launched online education.In online education,with the expansion of education scale and changes in teaching methods,the workload of teachers' assessment is increasing.Cumbersome,for the results submitted by students,the existing automatic scoring system can accurately discriminate objective questions such as multiple choice questions,fill-in-the-blank questions,and judgment questions.However,subjective questions mainly rely on manual scoring,which greatly limits the online The development of education.The implementation of automatic scoring of subjective questions in mathematics will greatly promote the personalization and intelligence of the intelligent educational guidance system,thereby improving the quality and effectiveness of educational services.At present,the automatic grading of elementary school mathematics application problems still faces three levels of technical bottlenecks,such as the existing automatic scoring method is not suitable for elementary school mathematics application problems,the existing scoring method has low accuracy,and how to automatically grade answers to multiple questions.In response to the ajor needs and challenges in the above three aspects,based on the text similarity calculation method and other theories,this paper has achieved the following innovative results in the study of the automatic scoring method of primary school mathematics application questions and the automatic scoring of multiple questions in one question:First,an improved method is proposed for automatic scoring of elementary school mathematics problems.For automatic text-based subjective question scoring,the sentence similarity calculation method is used as the foundation and the key technology.Its accuracy and efficiency are an important sign of the subjective question automatic scoring model,because this article is based on the application of primary school mathematics.Objects,compared with plain text type topics,have different characteristics of examining mathematical formulas,theorems,and mathematical logic analysis capabilities.Using simple sentence semantic similarity calculation methods can no longer meet the scoring requirements,so this article starts with similarity calculation algorithms.First Introduces the currently widely used and similarity calculation algorithm for subjective question automatic scoring,the principle of BLEU,and the research on automatic subjective question scoring that has been carried out at home and abroad,analyzes and summarizes their advantages and disadvantages,and aims at the application of primary school mathematics in China The characteristics of an improved semantic analysis-based sentence similarity calculation logic analysis scoring method and similarity calculation formula based on multi-feature fusion.The improved algorithm and formula not only consider the semantic information of the answer,but also solve the problem The logic sequence and key problem-solving steps are integrated into the calculation of similarity.This method can be better applied to the automatic scoring of mathematical subjective questions,and its discrimination accuracy is also higher.Second,two solutions are proposed for one problem with multiple solutions.For elementary school mathematics problems,there are often many ways to solve the problem,especially for higher-level problems,which is more common.For example,when solving the travel problem,either the hypothesis method or the quadratic equation method of one variable can be used.There may be multiple methods even for a certain step of the same problem-solving idea,so when scoring the paper,we must consider the test question scoring process of multiple standard answers.Before automatically scoring a multi-solvent elementary school mathematics algebra application question,we must first select the corresponding standard answer for the answers of different students of the same question.For the same question,different solutions will lead to solutions The problem ideas are different from the problem-solving steps.We can use the difference of the key steps in each problem-solving method as the characteristics to distinguish different problem-solving methods.This paper proposes two automatic scoring methods for solving one problem with multiple solutions,namely fuzzy matching method and S2 model method.In this way,even if the reference answer does not give all the solution methods in the automatic scoring process,the student's answer can be accurately scored.Third,design and implement an automatic scoring system model for subjective questions based on improved algorithms.This paper builds a subjective automatic scoring model based on an improved logic-based analysis method for elementary school mathematics application questions,and designs and implements each section of the scoring model.Through experiments on a large number of primary school mathematics problems,the results show that the improved scoring method and the designed automatic scoring model have certain practicability and superiority.In summary,this paper proposes an improved semantic analysis-based similarity calculation logic analysis scoring method and similarity calculation formula based on multi-feature fusion.This method can be better applied to the automatic scoring of mathematical subjective questions.The accuracy of its discrimination is also higher;two methods of fuzzy matching method and S2 model verification method are proposed for automatic scoring of multi-solution problems;a model of subjective question automatic scoring system is designed and implemented to verify the proposed method And the practicality and effctiveness of the algorithm. |
PDF Full Text Request