Research And Application Of Understanding Elementary Mathematical Problem

Recently years,human-like answer,machine certification,and natural language processing have become research hotspots with the development of artificial intelligence.The research and application of key technologies in artificial intelligence such as natural language processing in the field of mathematics can promote the construction of mathematical automatic reasoning platforms,it also has a certain significance for the intelligent education of elementary mathematics.This thesis studies the understanding of elementary mathematics problems,first analyzes the characteristics of elementary mathematics and uses predicate logic to express knowledge in order to ensure versatility,conciseness,and deductive reasoning accuracy for designing a complete knowledge architecture.The thesis proposes a method of comprehension of sentence meaning based on sentence patterns in Chinese Language.Firstly,a preprocessing process is established for mathematical text,the text is transformed into an abstract word sequence after word segmentation,part-of-speech tagging,named entity recognition,etc.Using finite automata to match word sequences with sentence patterns.When the sentence is matched successfully,specific information is extracted from the abstract entity tag,and the sentence is converted into the correct knowledge expression form through the normalized interface.Sentences with the same semantic meanings but different forms will fail to match if sentence patterns are not covered because the sentence model method adopts exact matching.Considering this problem,this thesis proposes a new method of comprehension of elementary mathematics problems based on the combination of entities as a supplement.This method uses the previous empirical knowledge form sentence pattern matching or artificial annotation to build a knowledge instance database.For sentences that fail to match the sentence pattern,the method will find the closest semantic sentence from the knowledge instance database and extract the mathematical entity information from the original sentence in order to combine the corresponding knowledge forms to complete the understanding of the questions.Semantic similarity is measured by the similarity of sentences.This thesis improves the deficiencies of previous similarity calculations by considering only a single feature,combines the characteristics of elementary mathematics sentences,improves the calculation method,and integrates semantic features based on word vectors and structural features based on editing distances.The performance of new improved similarity calculation method in terms of accuracy and recall rate has been improved through the experimental data.According to the process of understanding the meaning of the questions,this thesis designs and implements an elementary mathematics problem understanding system based on the above methods.The system runs stably on the comprehensive test.Combining the advantages of the two methods,the system has achieved good results in the accuracy and completeness of knowledge generation and has certain application value.