Dataset discovery
Describe your project goal in plain language or write a Boolean query. Then review dataset records with their source, stated license, and available metadata, and open each one at the repository that hosts it.
Free shows the top 5 datasets per search. Premium shows the full ranked list.
Overview
Assistemia's dataset finder helps researchers locate data for a specific topic, task, or modality without checking repositories one by one. It works as a dataset search engine for academic projects: describe what you need and let AI build the search strategy, or write the Boolean query yourself. Each result shows the repository it comes from, the license the source states, and available metadata such as format, size, sample information, author, and last update, with a Relevant or Partially Relevant label when one is assigned. It is most useful when you are planning a thesis chapter, a machine-learning experiment, or a replication study and need to know which candidate datasets exist before you check their access terms at the source.
Explore the Dataset Finder tutorialCapabilities
Describe the subject, data type, and task in 2 to 125 words, for example retinal fundus images for vessel segmentation. AI turns the description into a structured Boolean query and runs it straight away. The generated query is then available in Manual Search, so you can see how your request was interpreted and adjust the terms.
For full control, Manual Search accepts AND, OR, NOT, parentheses, exact phrases in quotes, and wildcards, and flags syntax problems before the search runs. An optional Last Updated year range narrows the records by update date. It suits a precise research dataset search where you already know your field's vocabulary and synonyms.
Narrow the loaded results by provider, domain, format, relevance, license type, license, size, data type, sample size, or last-updated year, or search within them, without running a new search. Cards show the source, stated license, and metadata the repository provides, and recognized licenses have a plain-language summary in the filter panel.
How it works
In AI-based Search, write a short project goal that names the subject, the kind of data, and the intended task. In Manual Search, enter a Boolean query and, if needed, a Last Updated year range. No file upload is required.
Results arrive as a ranked list with source, license, and metadata on each card, plus a relevance label where available. Filter by provider, format, or license to build a shortlist, and reopen earlier runs from Recent Searches.
Open a record to visit its repository page. Confirm the license, access conditions, version, file contents, and preferred citation there before you download, reuse, or cite the data, and keep a note of where each dataset came from.
Inputs, outputs & coverage
The tool helps you discover candidate datasets and compare their metadata. Decisions about access, licensing, and fitness for your study stay with you and the source repository.
Read our privacy policyPlans & limits
Try both search modes and the result filters. Each search shows its top 5 datasets, and monthly usage is limited; see Pricing for current terms.
Premium removes the Free results cap, so you can review the complete ranked result set a search returns and compare more candidate datasets before choosing one.
Frequently asked questions
Practical answers about sources, search modes, filters, and what to check before you use a dataset.
Contact our teamIt searches indexed dataset records from research repositories and data platforms, including sources such as Kaggle, Hugging Face, and Zenodo. Every result card names the repository it came from, and the Provider filter lets you keep only the sources you trust for your field. Coverage depends on what each repository exposes and on Assistemia's index, so a dataset that does not appear may still exist. If your discipline has a specialist archive, search it directly as well.
Use AI-based Search when you can describe the project but are unsure of the exact terms: write 2 to 125 words naming the subject, the type of data, and the task, and the AI builds and runs a Boolean query for you. Use Manual Search when you know your field's vocabulary and want full control with AND, OR, NOT, quotes, wildcards, and a Last Updated year range. You can start with AI and then refine the generated query manually.
Yes. After a search, the filter panel narrows the loaded results by provider, domain, format, relevance label, license type, specific license, size, data type, sample size, and last-updated year, and a search box finds text within the results. Filtering happens on the page, so it does not start a new search. For licenses the tool recognizes, an info icon gives a short plain-language summary, which is general guidance rather than legal advice.
No. It shows the license and metadata a source reports, but it does not verify them, host files, or grant access. Before using a dataset, open its repository record and confirm the license and any terms of use, whether access is open or by request, the version and files, and how the creators ask to be cited. Data involving human participants may also require ethics approval or a data-use agreement, so check with your institution.
Dataset Finder
Describe your project goal or write your own query, then review ranked dataset records and verify the best candidates at their source.