John Templeton Foundation · Philanthropy · Maestro
A classifier that sorts grants and papers by topic
The John Templeton Foundation funds interdisciplinary research. It accepts ideas on any topic but focuses on its priority areas.
| Area | Before | After |
|---|---|---|
| Classification | By hand, inconsistent | A tuned OpenAlex classifier |
| Data | Siloed | Grants and papers analyzed together |
The problem
- Industry
- Philanthropy
- Location
- West Conshohocken, Pennsylvania
JTF classified grant applications and academic publications by hand. The work was cumbersome and inconsistent, and it couldn’t scale.
Manual handling also risked data integrity and security. Data sat in silos, which limited what JTF could learn from its own grants and publications.
The usual alternatives fell short. Proprietary meta-databases were inadequate and expensive, and traditional machine learning needed extensive training data first.
Any new approach also had to work with JTF’s existing systems and workflows, and prove accurate and reliable under testing.
What we built
We started from OpenAlex’s text classification tool and customized it for JTF’s requirements and research focus areas.
We added relevant topics to the model and fine-tuned it to categorize grants and publications more accurately.
We connected the classifier to JTF’s existing systems and workflows through APIs, then tested its accuracy and reliability.
The result
Grant applications and publications are now classified more accurately and consistently, with less manual work.
With data no longer siloed, JTF can analyze grants and publications together and spot trends across the research it supports.
