Community-maintained directories and curated lists of datasets, meta-resources that help you discover other sources.
5sources
Curated lists are directories of datasets maintained by people rather than algorithms, often in the “awesome list” style you find on GitHub, where a community keeps a running, categorised index of the best resources in a field. For anyone building a RAG system, where the quality of your answers depends entirely on the quality of the knowledge base behind them, these lists are a shortcut through the discovery problem. Instead of guessing what exists, you start from a shortlist someone knowledgeable has already assembled.
The trade-off to weigh is freshness against breadth. A tightly focused list, say one devoted to legal data or natural language datasets, tends to be better maintained and better judged than a sprawling everything list, but it covers less ground. A broad list gives you more to explore at the cost of more dead ends. Check when the list was last updated and whether it is still actively maintained before you lean on it.
The main thing to watch out for is that a list is a signpost, not a guarantee. Link rot is common, so expect some entries to point at datasets that have moved or vanished. Curation rarely means vetting, so the presence of a source is not proof it is good or current. And crucially, these lists almost never carry licence information, so once a list leads you to a dataset, you still have to check that dataset’s own terms before using it.
The lists below are trusted starting points for finding data well beyond what we catalogue here.
A community-maintained list of legal datasets, tools, and resources for legal text processing across jurisdictions, including court records, statutes, contracts, and legal NLP benchmarks. A useful map for anyone building a legal RAG system.
A community-curated list of high-quality open datasets on GitHub, organised by topic: agriculture, biology, climate, economics, education, finance, government, healthcare, and more. A good starting point when you need RAG-ready data for a specific domain and do not yet know where to look.
A curated list of UK-focused open datasets from DataKind UK, covering government, health, crime, housing, and social data. A quick way into British public data when you are building a RAG system with a UK focus.
An alphabetical list of free and public domain text datasets for natural language processing, covering corpora, dialogue, sentiment, and summarisation. Handy when you want text-heavy data to build or evaluate a RAG system.
The UK's largest collection of economic, population, and social research data for teaching, learning, and public benefit. Many datasets need registration or an institutional login, so plan for an access step before you build with them.