Research On Readable Solution Generation Of Physics Circuit Problem For Intelligent Tutoring

With the wide application of computer technology in the field of education,the intelligentization of education has also developed rapidly,and various intelligent tutoring systems have begun to be applied to actual teaching services.As the core technology of the intelligent guidance system,machine answer can effectively improve the system's intelligence and the quality of teaching services.Both the teaching guidance and learning assessment of the intelligent tutoring system require a readable machine answer process.The project team proposed a problem understanding method based on relationship extraction to obtain quantitative relationship groups,and instantiate the relationship groups into equation groups,and solve the problems by solving the equation groups.On the basis of this research work,this thesis puts forward a method for generating readable answers to junior middle school physical circuit questions for intelligent guidance,and designs and implements a simple readable answer system.The main work of the paper is as follows:1)Analyzed the characteristics of readable solutions.The solution process of the junior middle school physical circuit problem is to start from the known conditions,combined with the junior middle school physical circuit related formulas and theorems,and gradually deduce the solution to the problem.It can be seen that the solution of circuit problems has a strict logic derivation process,and the previous step has a direct or indirect causal relationship with the next step.And describe the logic derivation process in the form of natural language.Therefore,readable answers include a rigorous logical derivation process and a readable description of this process.In order to generate readable answers,we convert this logically related problem-solving process into problem-solving action sequences,and then propose a readable description language model to generate readable descriptions of the solution actions.2)The method of generating action sequences is studied.The junior middle school physical circuit problem solution process is converted into an action sequence,and the action sequence for solving the problem is generated by constructing a sub-equation system.Through the machine solution method of relationship extraction,the problem-solving equations are obtained and used as candidate equations,multiple sub-equations are constructed through equation reasoning,and then correlation analysis is used to remove sub-equations that are not related to the problem.Finally,the remaining sub-equations are converted into solving actions,and a sequence of actions to solve the problem is output.3)A language model for readable description is proposed.Translate answer actions into content described in natural language.In the language model,the solution action is matched with the readable description template,and then the solution action is filled into the description template to generate a readable description.Finally,after the action sequence is reconstructed,it is translated step by step to generate a complete readable description answer.4)Designed and implemented a simple readable answer system.Specifically,the implementation method of the action sequence generation module and the readable description generation module is elaborated.40 test questions are constructed.Through the test,the correct rate of the generated action sequence is 100%,and the generated readable description is more interpretable.