Interdisciplinarity Inference Using a Cross Encoder between Research Projects

We propose an inference method that infers the feasibility of interdisciplinary research by integrating two existing research projects from the brief descriptions of those two projects and evaluated its performance. The proposed method utilizes a cross encoder that can classify various relationships between pairs of sentences as binary or multivalued. We demonstrated the effectiveness of the proposed method by experiments using the reports of the research projects conducted by the Grants-in-Aid for Scientific Research (KAKENHI). The proposed method achieved the F
1
score of 0.8318.