Key Technology And Application Of Understanding Elementary Mathematical Problem

In recent years, with the rapid development of information processing technology, Intelligent Tutoring Systems has become more and more concerned. To implement an Intelligent Tutoring System, an important prerequisite is how to make computers understand the mathematics problem that expressed in natural language. Then we need to convert mathematics natural language to formal knowledge representation which can be used for computer reasoning. Therefore, the study of understanding elementary mathematics problem is important. In this thesis, the main work is divided into two parts: the research of general mathematical problems comprehension and mathematical word problems(probability problems) comprehension.In understanding of general mathematics problem, we analyze the concise, logical, versatility characteristics of mathematics problems, then propose a set of knowledge representation methods based on predicate logic and sentence models. Using machine learning theory and artificial annotated corpus to train the models for recognizing entities of mathematics. After synonyms normalization treatment, a sentence is converted into a sequence. The basic unit of the sequence is word and its special symbol. The method of sentence model matching is based on finite state automata. After the sentence model matches the sentence successfully, it will extract significant information to generate knowledge representation, and then realize the general elementary mathematical problem understanding.Since various types of math word problems involved, we choose the probability word problem as the research problem. We research the characteristics and the probable word problem, propose an extended model based on mathematical probability problems by Kintsch. The model can effectively express objects, relationships between objects and problems. Probability word problems are a combination of several propositions together, different combinations constitute different situations. Refer way of understanding general elementary mathematics problem and further abstract word problems. Achieve a theory and method based on semantic sentence model to understand probability word problems.In a large number of general mathematics problems testing and analysis, the sentence model can correctly understand the meaning of the problem, when the sentence model completely covers the sentences problems. The main reason is the general mathematical problems are used more widely and universal. Therefore, the sentence model has a certain robustness. Probability word problems are more complex, so we need to write more sentence models for one situation, furthermore, we will increase and improve the model library to solve this problem better.