New Indicator of Science and Technology Inter-Relationship by Using Text Information of Research Articles and Patents in Japan

In this study, the text information of academic papers (about 2.3 million) published by Japanese authors and patents filed with the Japan Patent Office (about 12 million) since 1990) are used for analyzing the inter-relationship between science and technology. Specifically, a distributed representation vector using the title and abstract of each document is created, then neighboring documents to each are extracted using cosine similarity. A time trend and sector specific linkage of science and technology are identified by using the count of neighbor patents (papers) for each paper (patent). It is found that the number of papers with similar contents of patents decreased over time while the trend of patent counts with similar contents of paper is relatively stable. It is also found that the scope of scientific discipline by papers is relatively stable, while the technology fields by patents shows more dynamic patterns. This paper proposes a new methodology of measuring science and technology interlinkage by using textual information as a complement to traditional indicators based on non-patent literature citations of patents.