Research And Application Of Semantic Enhancement Technology For Elementary Mathematics Natural Language Understanding Based On Knowledge Graph

With the development of artificial intelligence technology,artificial intelligence has been widely used in various fields.my country is a big country in education,and elementary mathematics is the top priority of elementary education.The combination of artificial intelligence and education is the general trend.This thesis is based on an elementary mathematics human-like problem-solving system.The primary task of this system is to understand the meaning expressed in elementary mathematics questions.However,due to the imperfect understanding of natural language,the usability of the extracted knowledge is poor.Therefore,this thesis proposes a solution to this problem.Semantic enhancement technology to improve the usability of the knowledge map of elementary mathematics topics and ensure the integrity of the knowledge map of elementary mathematics.The main work of this thesis includes the following aspects:(1)All the mathematical entities and relationships that have been constructed in the elementary mathematics knowledge map are analyzed.The research abstracts the entities into two categories: independent entities and non-independent entities,and abstracts the relationships into three categories: ownership relationships,self-loop relationships,and verb relationships.The original knowledge map is improved to provide hierarchical entity relationship support for the subsequent semantic enhancement technology.(2)Based on the improved topic instance knowledge graph,a method of bottom-up abstraction of mathematical entities and top-down concretization of relationships is proposed.On the one hand,the existing topic instance knowledge graph with poor usability the structure is simpler and clearer.On the other hand,the relationship information originally hidden in the entity is mined to realize the relationship enhancement of the text semantics.(3)Furthermore,semantic analysis is performed on the expressions in elementary mathematics,and entity relationship information such as assignment units and naming units in the expressions are automatically extracted,so as to realize the relationship enhancement of the semantics of mathematical expressions.(4)Finally,this thesis applies the semantic enhancement technology to the humanlike answering system for elementary mathematics,and tested it on 700 elementary mathematics questions built by the research group.The average correctness rate of the human-like answering system for elementary mathematics increased by 20%.And the average problem-solving efficiency has also increased by 20%.