Approach of Intelligence Question-Answering System Based on Physical Fitness Knowledge Graph

Artificial intelligence’s penetrating sports is a new development trend of modern sports. Physical Intelligence Question-Answering System (QAS) is a typical application of artificial intelligence in sports, which can quickly respond the physic fitness questions raised by people. This paper aims the construction method of physical fitness QAS based on knowledge graph. Firstly, the physical fitness knowledge graph is constructed based on the crawling data and expert knowledge. Secondly, several physical knowledge question templates are constructed. Thirdly, the Bayesian classifier is used to classify the questions and the Bidirectional Long Short-Term Memory (BiLSTM) combined with Conditional Random Fields (CRF) method is applied to extract contents from the input questions. Then a matching algorithm based on bidirectional slicing string and a statistical method are performed to implement the fuzzy query to enhance the accuracy and robustness of the QAS. The experiments show that the accuracy of physical fitness QAS can reach 93.5% when the question sentences are matched with the query templates, and 86.5% when not matched.