Our knowledge-driven society demands more interplay between science and technology, and between research and industry. For example, by identifying links between authors of scientific publications and inventors of patents, more points of exchange between academic research and technological development are revealed.
In the paper "Measuring Industry-Science Links Through Inventor-Author Relations: A Profiling Method," authors Bruno Cassiman of IESE Business School and KU Leuven, Patrick Glenisson of KU Leuven and Business & Decision Benelux, and Bart Van Looy of K.U. Leuven and Steunpunt O&O Statistieken introduce a new profiling method to uncover existing links between authors and inventors.
Previous research shows that accounting for such links in the field of genetics increases the count patents related to university from 30 percent of patents with a university as assignee to 52 percent of patents with at least a university employee as inventor. A study on the field of nanotechnology and nanoscience shows that inventors who also publish are relatively more productive and more cited than the average scientist working in the field. Therefore, carefully accounting for these occurrences is a first step in stimulating these types of interactions.
Cassiman, Glenisson and Van Looy extend the current name-based approach through content-based lexical matching procedures. Their project is important because current name-based methods pose several important challenges, including false negatives and false positives, i.e. finding no link when one exists or finding links where none are present.
The authors demonstrate the capability of a pure lexical matching methodology. To achieve the connection between authors of scientific publications and inventors of patents, they link two different databases: the Web of Science (WOS by 151-Thomson), a publication database which features academic and scientific publications, and the database covering all the issued or granted patents of the European patenting system (EPO).
They apply the methodology to a specific case, that of the Interuniversity MicroElectronics Centre (IMEC) in Leuven, Belgium. IMEC is a leading research and development (R&D) center in the areas of microelectronics, nanotechnology, design methods and technologies for ICT systems. Its mission is to carry out R&D programs which are three to 10 years ahead of today's industrial needs through European and worldwide collaboration with industry and academia. The semiconductor industry is suitable because it is known for its lively interaction between science and industry.
The team's first step was to map and couple activity profiles of authors and inventors by creating content-based matches - between authors and inventors - through similarity between the (text) profiles derived from patent documents and articles, respectively. Next, they compared inventor and author names on the highest ranked matches for the occurrence of name matches. Finally, they compared the candidate matches with the names listed in a validated set of inventor-author names.
"Our text-based profile methodology performs significantly better than a random matching of patents and publications, suggesting that text-based profiling is a valuable complementary tool to the name searches used in previous studies," they write.
A key improvement of their match making process is the use of text-mining techniques, which goes beyond the simple coupling of authors and inventors based on their first and last names. They use the information contained in publication and patent abstracts to generate keywords to associate the author of a publication and the inventor of a patent. The novelty of this technique consists in generating keywords through text-mining rather than through the more standard use of search-strings which requires that the researcher pre-specify what kind of information he or she is looking for.
The ultimate goal of the pure lexical matching methodology is to use it in conjunction with name searches by eliminating false positive matches or by generating additional candidate matches that existing matching methodologies based on name searches might have missed.