How to Use Dataset Finder
Discover research datasets from Kaggle, Hugging Face, Zenodo, and 6 global repositories
Step-by-Step Guide
Choose Your Search Mode
Use AI-based Search to describe your dataset need in plain language, or Manual Search for full control over your search terms with AND, OR, NOT, quotes, and wildcards.
Enter Your Dataset Need
In AI-based Search, describe the subject, modality, and task together. In Manual Search, enter your search terms directly and optionally filter by Last Updated year range.
Search Across Global Repositories
Dataset Finder searches hundreds of thousands of indexed datasets from Kaggle, Hugging Face, Zenodo, Figshare, Mendeley, and more to find the most relevant matches for your research.
Review the Ranked Results
Results are ranked and presented with source, description, format, size, sample information, author, and update date. When available, datasets are labeled as Relevant or Partially Relevant to help you prioritize.
Refine Results on the Page
Use the result-page filters to narrow the loaded result set by provider, domain, format, relevance, license, size, data type, sample size, and update year. You can also search within the loaded results without rerunning the backend search.
Tips & Best Practices
- In AI mode, describe the subject, modality, and task together for the most relevant results
- In Manual Search, group synonyms with OR and connect major concepts with AND
- Use NOT sparingly because overly strict exclusions can remove useful datasets
- Apply Provider, Format, and Relevance filters after the search instead of overconstraining the initial query
- Check both Relevant and Partially Relevant rows when recall matters more than strict precision