Hierarchical Interdisciplinary Topic Detection Model for Research Proposal Classification

The peer merit review of research proposals has been the major mechanism to decide grant awards. However, have become increasingly interdisciplinary. It a longstanding challenge assign interdisciplinary appropriate reviewers so are fairly evaluated. One critical steps in reviewer assignment is generate accurate topic labels for proposal-reviewer matching. Existing systems mainly collect manually generated by principle investigators. such human-reported can be non-accurate, incomplete, labor intensive, and time costly. What role AI play developing fair precise proposal system? In this study, we collaborate with National Science Foundation China address task automated path detection. For purpose, develop deep Hierarchical Interdisciplinary Research Proposal Classification Network (HIRPCN). Specifically, first propose hierarchical transformer extract textual semantic information proposals. We then design an graph leverage GNNs learn representations each discipline order knowledge. After extracting knowledge, level-wise prediction component fuse two types knowledge detect paths proposal. conduct extensive experiments expert evaluations on three real-world datasets demonstrate effectiveness our proposed model.