Improving The Accuracies Of Word Segmentation And POS Classification Of Arithmetic Word Problems

With the development of Internet technology, artificial intelligence has received extensive attention and application in the field of education. It is also a hot topic to help students improve problem solving ability by machine answering mathematic application problems. From a wider range,the development and application of education robot has greatly improved the scientific literacy of students, and it has made many fruitful results in teaching field, which attracted more and more attention.The understanding of the problem is an important step for machine to solve the math problem, and the Chinese word segmentation and POS technology play a key role in the understanding of the subject. However, Chinese word segmentation and POS technology,which have a good application in many fields,have lots of problems in dealing with mathematic. The research that improving the accuracy of word segmentation and POS is put forward, in order to lay a solid foundation for the machine solving the mathematic application problem. This paper follow the latest mathematic problem-solving framework,which creates a set of semantic models, extracts quantitative relationship of the mathematic application problem based on the matching template, and form equations to achieve the understanding of mathematic problem. Under this technique, the research is based on the matching template that created in the process of understanding mathematic problem for machine. Through classification statistics, annotate and modify the problem of POS tagging in mathematics application problem, improving the accuracy of the word segmentation and POS for the machine to solve the primary school mathematics application problem.The main contents of this paper are summarized as follows:Firstly, the PEP primary school mathematics application problems is segmented by the NLPIR, which is Chinese word segmentation system, for the part of speech tagging,and summarizes the scores, Chinese units and other 9 categories of POS tagging problem by analyzing the problem of POS tagging which does not match the matching template.Secondly, designs the method for improving the accuracy of segmentation and POS tagging problem, and achieve the accuracy rate more than 90%, which effectively improves the accuracy of POS tagging. Finally, adding the primary school mathematics (Jiangsu Version) and Excellent Students Textbooks as a test, through the experimental statistical analysis,the method are verified effective and applicable, and it provides a technical basis for better creation of semantic models.