Classifying software citations

[bmcfee]

The mapping between paper text/citations and software projects will probably be difficult. Here are a few ways I see this going down:

• Direct citation of software with a proper DOI. This our best case.
• Citation/footnote of software via url. This is probably more common (depending on field), but also not hard to catch.
• Citation of a paper related to a project. Some software authors prefer this method, since it maps directly onto their h-index. Downside: these look just like ordinary paper citations. We'll probably need a preliminary web crawl of projects in this category before analyzing paper text.

What else can go wrong? Have other previous studies run into and/or solved these issues already?

[bmcfee]

Oh yeah, the worst (detectable) case:

• Software mentioned in plain text with no url, DOI, or direct attribution.

[rjleveque]

There's a related page at https://github.com/uwescience/citing_software that I started at a PDE software conference last summer that contains some links of possible interest.