Patent technology plays an important role in knowledge transfer between academia and industry. Previous methods use databases and search engines to help industrial users find suitable academic patents. However, academic patents are scattered in different universities and previous methods react to users’ needs passively. In the era of big data, there is a need of efficient channels where companies can easily learn academic patents and researchers can easily promote their patents. Besides, there is a need of proactive means to help companies find suitable academic patents from a large number of potential choices. Therefore, this research proposes a knowledge graph approach to patent recommendation based on an online knowledge transfer platform.

The knowledge transfer platform (i.e., Technology Marketplace of Jiangxi Province, R.P. China) is built to connect researchers to company users for the knowledge transfer. Based on the platform, the proposed knowledge graph approach recommends suitable academic patents to company users. The proposed approach differs from previous approaches mainly in profiling academic patents and companies, and capturing different needs of the companies. More concretely, the proposed approach defines and constructs a patent knowledge graph to capture the semantic information between keywords in the patent domain. It then profiles patents and companies as semantic graphs based on the patent knowledge graph. Finally, it generates recommendations by comparing the graphs based on a graph edit distance measure. During the recommendation process, three recommendation strategies (i.e., supplementary, complementary, and hybrid recommendation strategies) are proposed to meet different needs of companies. The supplementary strategy recommends patents that are similar to the patents of a target company. The complementary strategy recommends patents similar to the patents that are popular in the associated industry but are not owned by the target company. The hybrid strategy learns companies’ preferences for supplementary and complementary patents, and generates recommendations accordingly.

The proposed knowledge graph approach has been implemented and tested on the knowledge transfer platform in Jiangxi province, R.P. China. An offline experiment shows that the proposed approach achieves better recommendation performance than several baseline methods in terms of precision, recall, F-score and mean average precision. User feedbacks from an online experiment further demonstrate the usability and the effectiveness of the proposed approach for knowledge transfer between academia and industry.

This research provides a new approach and a practical solution to facilitate knowledge transfer between academia and industry. From researchers’ perspective, this research helps them promote their patents and may help them earn money. From companies’ perspective, this research helps them find suitable patents and may lower their searching cost. Besides, this research also responses to the calls from governments for facilitating knowledge transfer between academia and industry. The proposed approach can be applied in the scenarios where companies and their associated industries can be identified, and may be adapted to other university-industry collaboration applications, such as expert recommendation and project recommendation.