software-native metrics: depsy lessons learned

40
#3amconf Sept 2016 Heather Piwowar and Jason Priem @ImpactStory software-native metrics

Upload: heather-piwowar

Post on 21-Jan-2017

186 views

Category:

Science


2 download

TRANSCRIPT

#3amconfSept2016

HeatherPiwowarandJasonPriem@ImpactStory

software-native metrics

depsy.org

10.1038/529115a

depsy.org

launched last November

....what did people think?

some areas we weren’t sure about, but feedback suggested we got right

public profiles for everybody!

an overall score, with equal weighting to 3 components

downloads, log transformed vs ranked, and display for interprebility, with percentiles

pagerankto weight or not to weight, displayed in interpretable way

citations

authorship weighting (committers vs pypi author vs maintainer)

and some big requests

biggest one: pull in new data from GitHub

expected:

more package library systems (esp bioconductor)

more languages (import detection complexity)

more platforms like bitbucket (esp to get commit history for authorship)

better deduplication (no orcid, alas)

things that aren't on a version control system at all (not happening)

outstanding challenge:

deciding what is a research package (important b/c refset matters)

future plans

above, plus better text mining

integrate better with libraries.io

tighter integration with Impactstory

https://github.com/Impactstory/depsy-research/blob/master/introducing_depsy.md

more Depsy details:

thoughts?

tweet @depsy_org or via github issue.

thanks!