Abstract: For the purpose of improving the adaptability of the query reasoning system to the highly dynamic and rapidly expanding knowledge graph returning more efficient, real-time and accurate results to the users, a reasoning algorithm based on semantic tensor is proposed through comparative study of the prevailing knowledge graph reasoning algorithms. Two open network knowledge graphs were selected and simplified to be tested and trained by this algorithm. The experimental results show that this algorithm can adapt to the highly dynamic and consistently evolving data and information of the knowledge graph, improve the accuracy of the reasoning, intelligence of the search engine, and save the memory.

Key words: knowledge graph, semantic tensor, dimension reduction, matrix factorization, reasoning