The Long Windy Path to Visualising Science

Hamish
Litmaps
Published in
3 min readApr 20, 2021

We started Litmaps as a way to take the drudgery out of reading academic papers, and create a fast track to understanding the scientific landscape. No more copy-pasting titles out of reference lists and into Google Scholar. Replace wall-of-text literature reviews with beautiful visualisations of the key literature in the field. No more ad-hoc use of reference managers, spreadsheets, and text documents to keep track of the papers you’ve read or are relevant to you. Rather than finding the most important papers in your discipline by random walking through the citation graph, why not pre-compute the random walk and bring the most important papers directly to you?

Like everything in life, this ambition turned out to be a labyrinth of hidden complexities, trade-offs, and ambiguities. How can you visualise hundreds of citation connections without it being a horrible mess? Should already famous papers be promoted to users, or would that engender unfair Matthew Effects? How do you display fifty papers from the last ten years with less than ten citations alongside ten papers which are more than fifty years old have have ten thousand citations?

Litmaps 1.0 was a map of the entirety of scientific literature. This turned out to not be very useful: users are more interested in specific niches than science as a whole. So we switched to letting users incrementally build up a map from nothing by indicating papers they are interested in.

However, Litmaps 2.0 had some problems:

  • We weren’t sure how users would want to organise their papers, so we ended up with a grab bag of tags, maps, and read-later flags.
  • It was hard to get started if you wanted to bootstrap from a few papers that you knew of.
  • We also only included academic documents that had DOIs to keep out papers of ill repute. But this excluded important patents, preprints, and e-prints from Litmaps.

We’re excited to announce that we’ve just finished building Litmaps 3.0, which has fixed all these problems and more!

Here are the highlights:

  • We’ve pared down the way your papers are organised. You now have a set of maps, and each map is a set of papers. You can easily navigate between maps, add one or more papers to one or more maps, and investigate the intersection and interaction between different maps.
  • The new interface is also bright and colourful and will fill you with joy. Each map is a different colour which helps you visually separate out different topics within the same visualisation.
  • Creating a map is way easier, with new options like creating a new map from a seed article. You can even share an article as a seed with others!
  • We now have distinct workflows for quick search, explore, and (coming soon) systematic search (follow us on Twitter to find out when this is released). Explore in particular is great for iterating toward specific topics, as you add more items to your maps and winnow down results.
  • We’ve added all the non DOI items from Microsoft Academic Graph, adding those important missing patents, preprints, and e-prints from the last version. In the future we are looking at navigating between papers by authors, institutions, geography, and more (get in touch if this speaks to you).

With these updates, we think that Litmaps is well and truly the best place to explore, aggregate, and keep up to date with scientific literature.

We hope you like it!

--

--