Determining Research Priorities for Astronomy Using Machine Learning

TL;DR

This study explores using machine learning, specifically Latent Dirichlet Allocation on astronomy abstracts, to predict future research trends and priorities, potentially aiding strategic planning in astronomy.

Contribution

It demonstrates that LDA-based topic metrics from abstracts correlate with high-priority research areas and growth, offering a new tool for research trend prediction.

Findings

LDA topic metrics correlate with decadal survey priorities.

Growth rate of topics indicates emerging research areas.

Abstract-based metrics can predict future research interest.

Abstract

We summarize the first exploratory investigation into whether Machine Learning techniques can augment science strategic planning. We find that an approach based on Latent Dirichlet Allocation using abstracts drawn from high impact astronomy journals may provide a leading indicator of future interest in a research topic. We show two topic metrics that correlate well with the high-priority research areas identified by the 2010 National Academies' Astronomy and Astrophysics Decadal Survey science frontier panels. One metric is based on a sum of the fractional contribution to each topic by all scientific papers ("counts") while the other is the Compound Annual Growth Rate of these counts. These same metrics also show the same degree of correlation with the whitepapers submitted to the same Decadal Survey. Our results suggest that the Decadal Survey may under-emphasize fast growing research. A preliminary version of our work was presented by Thronson et al. 2021.