Abstract
Growing developments in general semantic networks, knowledge graphs, and ontology databases and the lack of such a resource for engineering and technology innovation have provided the motivation to build a large-scale comprehensive semantic network of technology-related concepts for artificial intelligence applications toward engineering design and innovation. Specifically, a Technology Semantic Network (TechNet) is constructed, covering the technical terms in all domains of engineering and technology and their semantic associations, by mining the complete U.S. patent database since 1976. To derive TechNet, natural language processing techniques were utilized to extract terms from massive patent texts and word embedding algorithms were employed to vectorize such terms and establish their semantic distance in a vector space, forging a semantic network. TechNet outperforms other semantic networks when evaluated for retrieving terms and their pairwise relevance that is meaningful from a technology and engineering design perspective. The semantic network of technical terms, as well as their latent embedding vector space, is a nearly decomposable semantic representation of the total technological knowledge space. A longitudinal analysis of TechNet using graph and information theoretic metrics reveals the fundamental structure and changes of the technological knowledge space. Several methods and tools based on TechNet for engineering knowledge discovery, topic mapping, precedents search, and concept generation and evaluation are developed. An online interface and APIs are created and made public to enable its access and utilization. TechNet is expected to serve as infrastructure for the development of a wide range of artificial intelligence capabilities and applications in the context of engineering and technology.