Related Experiment Video
Updated: Sep 13, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.3K
Automated Matchmaking of Researcher Biosketches and Funder Requests for Proposals Using Deep Neural Networks.
Sifei Han1, Russell Richie1,2, Lingyun Shi1
1Tsui Laboratory, Department of Biomedical and Health Informatics, Children's Hospital of Philadelphia, Philadelphia, PA 19146, USA.
IEEE Access : Practical Innovations, Open Solutions
|July 30, 2025
Summary
Researchers can now find suitable funding faster with a new automated system. This deep learning model accurately matches researcher profiles to grant opportunities, saving valuable time and improving success rates in grant applications.
Area of Science:
- Computer Science
- Information Science
- Computational Linguistics
Background:
- Researchers spend excessive time applying for grants due to low success rates and inefficient matching systems.
- Existing methods for matching researcher biosketches to funding opportunities often rely on unreliable keyword searches and generic recommendations.
- The U.S. faces a complex funding landscape with numerous federal and foundation grant programs, necessitating improved application efficiency.
Purpose of the Study:
- To develop and evaluate an automated matchmaking system for pairing researcher biosketches with funding requests for proposals (RFPs).
- To improve upon the limitations of current systems, such as keyword search unreliability and one-size-fits-all recommendations.
- To leverage advanced deep learning techniques for more precise and efficient research funding allocation.
Main Methods:
- Analysis of 12,991 researcher biosketches and 2,234 National Institutes of Health (NIH) RFPs from 2014-2019.
- Implementation and benchmarking of four deep learning models, including BERT with cross-encoding, Siamese networks, and BiLSTM layers.
- Comparison of deep learning models against traditional methods like logistic regression and support vector machines.
Main Results:
- The most effective model, integrating BERT with cross-encoding, a post-BERT BiLSTM layer, and back translation (BC2BT), achieved an F1-score of 71.15%.
- Advanced deep learning models significantly outperformed conventional predictive models in the biosketch-RFP matchmaking task.
- The developed system demonstrated enhanced precision in identifying suitable funding opportunities for researchers.
Conclusions:
- Sophisticated natural language processing (NLP) techniques, particularly deep learning, can automate complex matchmaking in the research funding sector.
- The BC2BT model offers a promising advancement for improving the accuracy and efficiency of grant application processes.
- This automated approach has the potential to streamline funding mechanisms and support future research endeavors.
