<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[Accidental Lawyers]]></title><description><![CDATA[Accidental Lawyers]]></description><link>https://accidental-lawyers.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Fri, 20 Nov 2026 16:09:21 GMT</lastBuildDate><atom:link href="https://accidental-lawyers.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[From CSV to citable: getting a DOI for your indie data project (and why every indie dev with public data should bother)]]></title><description><![CDATA[If you ship side projects that include any kind of public dataset — scraped, joined, cleaned, analyzed, anything — there's a 20-minute lever you're almost…]]></description><link>https://accidental-lawyers.hashnode.dev/from-csv-to-citable-getting-a-doi-for-your-indie-data-project-and-why-every-indie-dev-with-public-data-should-bother</link><guid isPermaLink="true">https://accidental-lawyers.hashnode.dev/from-csv-to-citable-getting-a-doi-for-your-indie-data-project-and-why-every-indie-dev-with-public-data-should-bother</guid><dc:creator><![CDATA[Marvin Bregiosa]]></dc:creator><pubDate>Mon, 01 Jun 2026 09:50:36 GMT</pubDate><content:encoded><![CDATA[<p>If you ship side projects that include any kind of public dataset — scraped, joined, cleaned, analyzed, anything — there's a 20-minute lever you're almost certainly not pulling: getting a permanent DOI for the data. Free. Indexed in Google Scholar. Citation-ready for any paper that ever mentions your work.</p>
<p>I just did this for <a href="https://accidentlawyerreview.com/research/vision-zero-report-card/">a 7-year analysis of US truck-fatality records I shipped last week</a>. DOI assigned in under 30 seconds after publishing the upload form. Cost: zero. Setup time: about 15 minutes.</p>
<p>Here's the walkthrough, plus the reasons it's worth the bother even if your project will never see an academic citation.</p>
<h2>What a DOI actually buys you</h2>
<p>A DOI (Digital Object Identifier) is a permanent URL-grade identifier that resolves through <code>doi.org</code> to your content, wherever that content lives. It's the citation primitive of academia, but it's not gated to academics. Two services issue them free for indie data:</p>
<ul>
<li><strong><a href="https://zenodo.org">Zenodo</a></strong> — operated by CERN, EU-funded, no friction</li>
<li><strong><a href="https://figshare.com">Figshare</a></strong> — commercial-side equivalent</li>
</ul>
<p>I went with Zenodo because the metadata model is more flexible and you can upload multiple files into one record (code + data + a methodology PDF, say) and the whole bundle gets one DOI.</p>
<p>The DOI you get back looks like <code>10.5281/zenodo.20487070</code>. It resolves to a Zenodo page that hosts your files, displays your metadata, and lets anyone download. The DOI itself is the durable identifier — even if Zenodo restructured its URLs tomorrow (it won't), the DOI keeps resolving.</p>
<h2>Why this matters even for non-academic indie projects</h2>
<p>Four things shift when your data has a DOI:</p>
<p><strong>1. Google Scholar indexes the record.</strong> Within a few days of publishing on Zenodo, your dataset shows up in Scholar search. That's a separate index from regular Google — a separate channel for discovery from researchers, journalists, policy analysts. Many of these people search Scholar first. For indie data projects this is a market segment with effectively zero competition from other indies.</p>
<p><strong>2. Schema.org gets a clean upgrade.</strong> If you ship your data on your own site too (you should), your <code>Dataset</code> JSON-LD now has a meaningful <code>identifier</code> field that points to the DOI. Google's Dataset Search uses this. So does data.gov's aggregator. So do academic crawlers. This single field meaningfully boosts how your data ranks in Dataset Search results.</p>
<p><strong>3. Citation gravity is real.</strong> When a journalist quotes a stat from your dataset, you've historically gotten a link if you were lucky. With a DOI, you become formally citable — which sounds like a small distinction until you notice it changes the behavior of the person writing about you. Many style guides actually require a DOI when one exists. Now you become the path of least resistance.</p>
<p><strong>4. Permanent attribution.</strong> Six years from now when your project shuts down and the domain expires, the data still lives at Zenodo with your name on it. That's actually nice. Indie projects die. Citations don't have to.</p>
<h2>The 15-minute walkthrough</h2>
<p>I'll assume you have a CSV (or JSON, or anything else under 50GB) sitting in a folder, plus some metadata in your head about what it is.</p>
<h3>Step 1: Sign up</h3>
<p>Go to <a href="https://zenodo.org">zenodo.org</a> and create an account. You can authenticate through GitHub, which I'd recommend because it auto-fills your name and links your profile. Zenodo respects the GitHub identity for the lifetime of your record.</p>
<h3>Step 2: New upload</h3>
<p>Click <strong>+ New upload</strong>. The form has a lot of fields but you only need to fill a handful.</p>
<p>The minimum required:</p>
<ul>
<li><strong>Resource type</strong>: Dataset (this matters — it controls how Zenodo and Scholar index)</li>
<li><strong>Title</strong>: Be specific. "US Truck-Involved Fatal Crashes (FARS) 2018-2024" beats "Truck data". Search engines parse this.</li>
<li><strong>Authors</strong>: You. Add affiliation if your project has a name — it becomes part of the citation.</li>
<li><strong>Description</strong>: Plain text. Cover what the data is, the temporal range, the source, and one or two methodology notes. Treat it like a paper abstract minus the academic register.</li>
<li><strong>License</strong>: CC BY 4.0 is the default I'd recommend. It requires attribution, allows commercial use, and is the license most journalists/researchers know.</li>
</ul>
<h3>Step 3: Upload files</h3>
<p>Drag in your CSV. If you have the script that produced it — and you should, for reproducibility — drag that in too. Both files become part of the same DOI'd record.</p>
<h3>Step 4: Related works (the under-appreciated step)</h3>
<p>This is the field most people skip and it's actually the highest-leverage part of the form. You can declare relationships between your Zenodo upload and other URLs:</p>
<ul>
<li><code>Is supplement to</code> → your project's main page (your own domain)</li>
<li><code>Is supplemented by</code> → the GitHub repo with the source code</li>
<li><code>Is identical to</code> → a HuggingFace mirror, if you have one</li>
<li><code>Continues</code> → an earlier version of the dataset on a previous DOI</li>
</ul>
<p>Each related work creates a bidirectional metadata link. Zenodo, Scholar, and Crossref all index these. If you've also mirrored to HuggingFace or hosted code on GitHub, declaring the relationships here is how each of those URLs gets cross-credit.</p>
<p>This is the indie equivalent of being part of an academic graph. Costs you 90 seconds. Pays dividends in discoverability for years.</p>
<h3>Step 5: Keywords</h3>
<p>Hand-roll 8-15 keywords. Zenodo's indexer is keyword-driven, so be specific and use the terms people in your space actually search. For my truck-fatality dataset I used FARS, NHTSA, Vision Zero, traffic safety, urban planning, data journalism, public data, and so on.</p>
<h3>Step 6: Publish</h3>
<p>Hit Publish. Zenodo mints the DOI immediately. You can then visit <code>https://doi.org/&lt;your-doi&gt;</code> and watch it resolve.</p>
<p>That's it.</p>
<h2>What I did right after</h2>
<p>Three things, in order:</p>
<p><strong>1. Updated my project's main page</strong> — added the DOI to the citation block on my own site. The DOI now appears as part of the formal citation, which makes it likelier that anyone copying the citation also propagates the DOI.</p>
<p><strong>2. Updated the GitHub README</strong> — added a "Cite this dataset" section near the top with the DOI badge (Zenodo provides an auto-generated SVG badge you can drop in).</p>
<p><strong>3. Mentioned the DOI in the HuggingFace dataset card</strong> — same record, same DOI, declared as a related identifier on the HuggingFace side.</p>
<p>This is the link-mesh pattern. The Zenodo record points at GitHub and HuggingFace; GitHub and HuggingFace point at Zenodo. Each platform indexes the others. Google sees a citation graph rather than three orphaned URLs.</p>
<h2>The honest caveats</h2>
<p>DOIs are not a silver bullet. They don't make bad data good. They don't substitute for actual reach. If your dataset is genuinely uninteresting, a DOI just makes the uninteresting more durable.</p>
<p>Two other things I'd watch:</p>
<ul>
<li><strong>You can't edit the files after publishing without minting a new DOI.</strong> Zenodo treats records as immutable. You can post a new version (which gets its own DOI, plus a "concept DOI" that aggregates all versions), but you can't silently fix typos. Make the dataset clean before you publish.</li>
<li><strong>CC BY requires attribution.</strong> That's exactly what you want for SEO and credit, but if you're hoping your data quietly becomes the foundation of someone else's commercial product without your name on it — wrong license. Use CC0 instead.</li>
</ul>
<h2>Why this is worth doing on a side project</h2>
<p>The pattern repeats across all the indie publishing channels I've used in the past two weeks for this same project — GitHub, HuggingFace, dev.to, Medium, LinkedIn, Zenodo. None of them require permission. None of them charge. Each one has its own discovery surface that doesn't overlap with the others. And every one of them lets a one-person project show up in places that look, from the outside, like the work of a much larger operation.</p>
<p>The DOI is the most specialized of the bunch and the one with the most academic gravity. It's also the one that takes the least effort. 15 minutes, free, permanent.</p>
<p>I'd do this for any indie data project, even ones with no academic angle whatsoever.</p>
<hr />
<p>The dataset I used as the example throughout: <a href="https://doi.org/10.5281/zenodo.20487070">US Truck-Involved Fatal Crashes (FARS) 2018–2024, on Zenodo</a>. The companion analysis and methodology: <a href="https://accidentlawyerreview.com/research/vision-zero-report-card/">Vision Zero Report Card 2026 on accidentlawyerreview.com</a>. The build write-up and bugs on <a href="https://dev.to/marvin_bregiosa/building-a-sortable-data-journalism-page-in-astro-33k-fars-records-sparklines-and-the-bugs-that-55i">dev.to</a>. The policy framing on <a href="https://medium.com/@arsen_26968/vision-zero-is-failing-in-8-of-17-us-cities-that-adopted-it-heres-the-data-62df264b7342">Medium</a>. The source on <a href="https://github.com/MarvinBregiosa/vision-zero-fars">GitHub</a>, the dataset mirror on <a href="https://huggingface.co/datasets/MarvinBregiosa/us-truck-fatalities-fars-2018-2024">HuggingFace</a>.</p>
]]></content:encoded></item></channel></rss>