Broader impacts NSF CAREER Award 1750978
This project seeks to make published (unstructured) evidence more useful by designing models that can automatically infer report findings. Meeting this aim would have substantial implications for the practice of evidence-based medicine, and more generally for scientific disciplines in which evidence (specifically, results of trials) is described in free-text reports.
As concrete examples of the impact these models might realize, we have translated research supported by this project by partnering with a few different groups. Specifically, we are working with [Cures Within Reach for Cancer] (https://www.cwr4c.org/blog/announcing-ai-partners) and researchers at Mass General Hospital. Both teams are using the methods developed under this project to help domain experts efficiently make sense of the vast literature base.
Education and research opportunities
As proposed in the grant application, I developed and executed the first implementation of Practical Neural Networks (Data Science 4440) as part of this project. I’ll be offering this again in the fall of 2020.
This project has also provided key opportunities for research. Eric Lehman, who has been supported by an REU for this project, has made major contributions and was first author on our NAACL 2019 publication, which he presented (alongside Jay DeYoung) at the conference in June 2019. He has since co-authored an additional two publications related to this project. On my nomination, Eric was awarded Honorabl Mention for the Outstanding Undergraduate Researchers award given by Computing Research Association (CRA), largely on the strength of the research conducted for this project.
Additionally, PhD student Jay DeYoung joined this project in fall 2018. He has co-authored multiple publications relating to this work, and the project will serve as the basis for this thesis. DeYoung also served as a teaching assistant (TA) for the aforementioned Practical Neural Networks course in spring 2019, gaining teaching experience.
Finally, Joshua Oquendo — an undergraduate student at nearby Roxbury Community College — worked on this project as an REU over the summer in 2019, culminating his presenting the work at an RCC event.