Show HN: Litmaps – Visual Research Discovery Tool
app.litmaps.co
app.litmaps.co
Our first attempt to solve this was with a "map of all science" (from 1400-today, 100M+ articles, 1B citations), which we did build but it didn't end up solving the problem that well. We re-approached the problem by pairing a time-based citation visualization, with helpful search tools, and project (or map) state. Our visualization lets you quickly see how search results relate to each other (optionally with nodes sized by citation count). We currently have keyword search, and a network search (called Explore) that scans for highly connected articles up to two degrees away from your map(s) and/or keyword searches. Lastly the project state is nice because as you build up key articles, network search is more targeted, and you can opt into email updates if there are any newly published works that connect to your map as we update our dataset.
It's currently in early access so it's free to use, and you don't need to create an account to get started. You can dive right in and start finding relevant articles to what you are working on or are interested in. Keen to hear feedback on our work so far. Thanks.
Have you considered flagging paths for publications that get retracted? I've always thought that a 'this paper you're using has three retracted parent papers,' would be valuable.
Excited to see where you guys go with this!
One gripe: when asked for a paper title, I tried submitting a DOI link instead and it failed dismally. You could probably reduce server load for title lookup by allowing this as an alternative.
Also, I would like some sort of visual indicator to distinguish between works-cited-by a paper and works-which-cite the same paper. Just glancing at the map, it's not that obvious which is which from color or shape and they're mixed up in the default listing. When I find an interesting paper I often want to dig into its antecedents first and look at its influence afterward, but I don't see a clear visual representation of temporality here.
A wishlist:
1) To be able to share maps with fellow researchers, not as pdfs or bib files, but as a url to the map so that they can interact with it. I can see that a "share link" is possible for an article, but not for a custom map. Such a link could also be used to embed such maps on sites like how one can with plots from https://ourworldindata.org/
2) To be able to get more metadata like sponsor, country, etc. like https://www.lens.org does.
3) To be able to upload pdfs and annotate them. I can see why this is dicey, but if possible, would be great to see an integration with something like https://fermatslibrary.com/margins
> It's currently in early access so it's free to use, and you don't need to create an account to get started.
I know software developers need to eat and totally respect your choice in making it a paid service in the future, but I will confess a deep desire for such a thing to be affordable, if not free.
Take a look at an example map here: https://app.litmaps.co/maps?doi=10.1038/S42256-019-0086-4
After clicking through all the links on the top navigation bar as well as some of the links in the footer, and I didn't see anything except for marketing fluff and testimonials.
I was hoping for some screenshots some kind of "features" page, with a section for each major feature of Citationsy, and an in-depth explanation of how it compares to other tools, plus a screenshot of actually using that feature.
Did I miss something? If I did, please make it easier to find out what Citationsy actually does!
I also tried logging in, but your free trial also demands that I enter my credit card info. Big red flag, no thanks.
Imho this would give a much better overview over the topics and research itself.
With "author, year" format it feels like a tool to analyse citation networks and how people cite each other; more about the social aspect than about the content of the articles.
And I'd ideally like to also see the abstract then, somewhere. Rather than that requiring a click. Currently, "surfing", orienting exploration, seems regrettably high friction. Click, wait-for-load, look, click, load, look, "hmm, did I click on that one already? yes, drat.", ... "have I clicked on all of them yet? I don't know, maybe I've missed something nifty, but I'm done.". Having faded pink nodes that no longer respond to mouse adds a "ok, to see that node, I currently need to first click on other node".
For node labels, perhaps have some kind of title digest? To visually aid the "remind me again which paper that node is?". Perhaps the title in an unreadably small font?
When "Run Explore" on a map yields "No results", perhaps add an onboarding hint of what needs to be different in order to have results?
Sigh. When using HN, after logging out in a different window, continuing to edit a comment, and clicking "update"... apparently unrecoverably discards the edit?!? :/
Briefly, on the map list of papers, the scrollbar vanishes after a second or two (on desktop firefox), so you lose the visual indication of how far down the list you are. Have to jiggle the list to briefly get that back.
On the map list of papers, clicking on a paper popovers the paper info. To go back, there's a close-X, and history-back works. But ESC doesn't, and perhaps should.
Off topic but for fun, when using a shallow-3D display (eg AR glasses), it seems information might be segregated/decluttered using depth. So for silly example, there might be lots of title clutter, but if it was on a layer raised above the node graph, one might "look through" the clutter to the graph. And then attentionally shift to the title layer, without that depth step being so large as to trigger refocus eye strain. ... Maybe. Just something I've played with. And it's less than pretty.
Sadly however, it seems that the .bib file I get from Zotero isn't supported, since it claims to have found 0 articles in it, despite the fact that I have verified those articles show up when searching for them directly.
Edit: TIL that you can also get a .bib file from the "export items" button, as well as from "create bibliography from items". I had been using the latter. It seems that Litmaps only works with .bib files created using "export items".
Where are you indexing your articles from? Plenty of ML/CS articles, but wasn't able to find someone by their Orchid ID in the Biology/Chemistry space.
Looking forward to the full release.
1. This is amazing. I wish I had it years ago, and I could use it now.
2. I like the visualization but my favorite thing is the sense of completeness and influence I get when I look at articles. Rather than relying on someone's opinion of the most influential article, I can see the article and the influence right there.
3. This is early days but I am already running into the frustration of not being able to see all the articles out there. This probably has to do with publishers and paywalls and what not, so probably not your fault, but it is still a problem that has to be dealt with in order to render this truly useful. ()
() interestingly, over time, it may not matter as much when every paper has a version on ARXIV, but in 2021 I was looking for papers influenced by work done as far ago as the 80s and 90s.