Abstract: Identifying and analyzing the domain technical competition situation is of great significance for mastering the overall development of science and technology and exploring the potential path of scientific and technological innovation. This article uses natural language processing technology to conduct text mining on patent abstracts, calculates the semantic similarity between patents, and uses eigendecomposition to extract competing patents in important positions to construct a competing patent network. Multi-dimensional analysis is executed based on competing patent networks. The results show that the competitive patent network in this field is fragmented; the distributions of patent institutions and domestic holders are not balanced; the technological competition between applicants is manifested in different scopes and different technical fields.

Key words: Text mining, Patent network, Competitive patents, Technical competition, Competition distribution