TechNet is a very large technology semantic network. It consists of over 4 million technology-related terms and pairwise semantic relations between them. The terms are retrieved from unstructured texts in over 6 million U.S. patents from 1974 to date. These terms represent technological concepts regarding functions, components, materials, configurations or working mechanisms. Their pairwise semantic relations are calculated using a word embedding model and indicate the technical relevance between technological concepts. TechNet outperforms Google Knowledge Graph, WordNet and ConceptNet for knowledge retrieval, inference and represention tasks in the context of engineering and technology (1,2). TechNet is a digital knowledge infrastructure to support a wide range of engineering knowledge-based AI applications, technical language processing and engineering text analysis (3, 4, 5, 6).
TechNet is freely accessible through open API. TechNet word embeddings are downloadable here.
Please refer to the following paper: Sarica, S., Luo, J. and Wood, K.L., 2020. TechNet: Technology semantic network based on patent data. Expert Systems with Applications, 142, 112995.
TechNet is created and maintained by Data-Driven Innovation Lab.