Roman Name Attestations puts around 250,000 Roman Empire inscriptions from the Epigraphic Database Clauss-Slaby (EDCS) into a single searchable interface. An AI pipeline extracts praenomen, nomen, cognomen, social status, and gender for the people named in each inscription. The whole thing is free, released under CC BY 4.0.
The question for Roman Name Attestations is whether it beats the standard workflow of cross-referencing EDCS by hand. For bulk work, yes. For a specific citation you plan to publish, verify by hand. The 80 to 85 percent extraction accuracy claimed on the site is honest, and it matters.
Key Features
The interface centers on an interactive map. Every dot is an inscription with a location, and you can filter by name component, gender, date range, deity or imperial status, and confidence level. The confidence filter is the one to know about: it exposes how sure the AI pipeline is about a given extraction, so you can trade recall for precision when the question demands it.
- Interactive geographic map covering the full corpus of around 250,000 inscriptions from EDCS.
- Per-individual extraction of praenomen, nomen, cognomen, social status, and gender.
- Search and browse across names, provinces, and dates.
- CSV and JSON export for any filtered result set.
- Inscription translations and AI-generated summaries where those exist.
- Community flagging so users can report a bad extraction and route it back to the pipeline maintainers.
What the export actually gives you
CSV and JSON are the two supported formats. Each row is one individual person rather than one inscription, which is the right shape for prosopographical work. If an inscription mentions four people, you get four rows keyed back to the same source ID. There is no documented API, so any downstream tool has to consume the exports. That works for a one-off study but limits any live application.
Pricing Breakdown
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Full map, search, filters, CSV and JSON export, community flagging |
There is one tier. Everything is free, the derived data is CC BY 4.0, and there is no paywall on any feature. The source corpus (EDCS itself) is also open, so nothing here is locked behind a subscription elsewhere.
Pros
- Scale. A quarter-million inscriptions in one queryable place is a real capability that did not previously exist in this form.
- Free and openly licensed. CC BY 4.0 on derived data means you can put results in a paper, a dataset, or a downstream tool without permission wrangling.
- The AI extraction, imperfect as it is, saves the kind of manual coding that used to eat months.
- Working exports. CSV and JSON both come out clean and match the structure shown in the UI.
Cons
- Name extraction sits at 80 to 85 percent accuracy. For bulk statistics that is workable. For a citation in a monograph it means every claim has to be checked against the original inscription text.
- The audience is narrow. If you are not doing Roman prosopography, Latin epigraphy, or adjacent digital humanities work, there is very little reason to open the site.
- No API. Everything happens through the UI, so a build that wants live queries has to scrape or ship exported snapshots.
- The site lives on new.roman-names.com. A subdomain prefixed with new usually signals a project mid-migration, and there is no public roadmap saying when the canonical URL flips.
Who Is It For
Classicists, ancient historians, and digital humanities researchers working on the Roman world get the most out of this. If you are running a study on Roman naming patterns, provincial identity, or imperial cult participation, this is the kind of shortcut that did not exist five years ago.
Hobbyist historians will get value from the map alone. The confidence filter, date range, and province picker let you do surprising things with an idle afternoon. Whether that justifies a bookmark depends on how often you find yourself wondering how many Marci Aurelii ended up in Britannia.
Verdict
Roman Name Attestations is a serious research tool priced at zero. Use it as the first pass. Run your filters, pull the CSV, do your bulk analysis, then verify every individual claim you plan to publish against the underlying inscription. The 80 to 85 percent accuracy floor is not a dealbreaker for statistical work, and it is disclosed openly, which is more than most AI-augmented databases will do.
My recommendation for anyone working on Roman-era material: bookmark it, read the confidence numbers on any extraction you cite, file community flags on the errors you spot (the pipeline appears to iterate on them), and keep EDCS itself in the loop for citation-grade work. Rating: 6 out of 10.