Alex D Wade (@alexwade) 19 September 2019
CZI Science, Open Science, & Meta
Presentation Name Date
Supporting science and technology
that will make it possible to
cure, prevent, or manage all
diseases by the end of the century
Accelerating biomedical science
by developing new tools and
technologies and supporting open,
collaborative models of research
10 years81 years
program officer: dario taraborelli
Projects request between
$50K and $250K for one year
Not limited to writing code
Docs, management, and
— the “invisible work” —
can all be supported!
graphic by @deniseyu21 from a talk by karthik ram (@_inundata)
+ full text (Publishers)
+ full text (crawled)
- Cross-Entity Recs
Meta Knowledge Graph
● Eigenfactor (EF) originally developed to assess journal importance.
● Extended to article-level (ALEF)
● Meta calculates an EF value
for every node in the graph
Bergstrom (2007); West el al (2010); Wesley-Smith, West, Bergstrom (2016)
Our approach: Predict the impact for new papers
● Problem: recent papers have zero/few citation edges so EF=0 but our users engage
almost exclusively with these new (uncited) papers
● Solution: infer an Eigenfactor value for recent papers
○ ML model trained on prior publications citations at 3 years
○ Validated on withheld dataset
● Predicted Eigenfactor (EFP) calculated on all new papers
○ EFP values used to rank/sort papers in feeds
○ EFP replaced with proper EF over time as citations begin to accrue
Meta: next steps
● Meta pubic launch: soon!
● We would love feedback & suggestions!
● Please come curate & share your own feeds!
● How can we partner with the community?