F-UJI

F-UJI checks dataset metadata against FAIR requirements.

Supported versions

Full metric support means all its checks are supported; partial means only some. A supported check still needs the required information to produce a decision.

Select a version explicitly to keep using its mappings, checks and scoring rules:

from fair_offline_assessor import Assessor

assessor = Assessor("FUJI", version="3.5.1")

What it checks

F-UJI looks for identifiers, descriptive information, licences, access conditions, links to data, relationships between datasets, and information about how data was created. Some checks examine whether recognised standards and file formats are used.

Each check has its own requirements and points. These examples use F-UJI 3.5.1:

Field names depend on the metadata standard you use. F-UJI's mappings connect those fields to the information each check requires. For example, Schema.org name supplies the dataset title.

The complete requirements and points are in F-UJI's check definitions for this version.

What can run offline

The seven unsupported checks remain indeterminate with reason unsupported_check.

Supported checks may also remain indeterminate if the required information is missing or cannot be interpreted. Offline results cannot confirm that a link opens, a dataset is downloadable or a search engine can find it.

Metadata

Pass document contents to metadata. JSON-LD is the default. For another format, set metadata_format to one of the values below.

DataCite JSON requires agency. DataCite API responses with metadata inside data.attributes are not supported. XML must contain its own information, without DOCTYPE or entity declarations. If an OAI-PMH XML response contains several records, F-UJI uses the first. RSS/GeoRSS and OAI-ORE Atom are not supported.

Supply one intended dataset per document. F-UJI decides which described dataset to assess using its own rules. With JSON-LD or other RDF formats, you can set subject to the identifier you expect it to choose. A choice that cannot be confirmed produces subject_not_selected. Omit subject for XML, DataCite JSON and HTML.

For Schema.org metadata:

• Use identifier for the dataset identifier required by the citation check. F-UJI does not use @id for that field.
• Describe creators as Person or Organization objects, for example "creator": {"@type": "Person", "name": "Ada Example"}. F-UJI 3.5.1 does not use a plain text creator value in RDF input.
• isAccessibleForFree: false is not used by this version's RDF mapping. That value alone cannot establish access restrictions.

from fair_offline_assessor import Assessor

result = Assessor("FUJI", version="3.5.1").assess(
    metadata={
        "@context": {
            "dcat": "http://www.w3.org/ns/dcat#",
            "dct": "http://purl.org/dc/terms/",
        },
        "@type": "dcat:Dataset",
        "@id": "https://example.org/datasets/1",
        "dct:title": "Example dataset",
        "dct:license": {
            "@id": "https://creativecommons.org/licenses/by/4.0/"
        },
    },
    metadata_url="https://example.org/datasets/1",
)

Schema.org, DCAT and Dublin Core are vocabularies: sets of named properties with agreed meanings. DataCite defines a metadata schema: fields and rules for describing a dataset. RDF describes things through statements such as “this dataset has this title”. JSON-LD and Turtle are ways to write those statements.

Contexts

A JSON-LD @context defines what field names mean. The library includes the Schema.org context, so "@context": "https://schema.org" works offline.

For another context URL, supply its document in local_contexts, or put the field definitions directly inside @context. Any other contexts it refers to must also be supplied. Context definitions cannot override different definitions already included in the library.

from fair_offline_assessor import Assessor

result = Assessor("FUJI", version="3.5.1").assess(
    metadata={
        "@context": "https://example.org/context",
        "@type": "Dataset",
        "name": "Example dataset",
    },
    local_contexts={
        "https://example.org/context": {
            "@context": {
                "@vocab": "https://schema.org/"
            }
        }
    },
)

Results

The common response includes scores for individual checks and metrics. The evidence entries refer to your whole metadata document or metadata_url.

raw contains unmodified F-UJI result dictionaries, with at most one entry per metric. Metrics that produce no result are omitted. It covers this offline run, not a full response from F-UJI's online service.

See Results for field meanings and Troubleshooting for problems with metadata.

Troubleshooting

Start with result.diagnostics for problems with the metadata. For a particular check, read its reason_code and message.

indeterminate means the library cannot decide whether the check passes. Other checks can still run if they have the information they need.

Example: a missing context

from fair_offline_assessor import Assessor

result = Assessor("FUJI", version="3.5.1").assess(
    metadata={
        "@context": "https://example.org/context",
        "@type": "Dataset",
    }
)

print(result.diagnostics[0].model_dump_json(indent=2))

{
  "code": "unknown_context",
  "message": "Context unavailable locally: https://example.org/context",
  "location": "/metadata"
}

Problems with metadata

Codes such as invalid_subject, invalid_metadata_url and invalid_local_contexts mean an argument has the wrong type or value. Check the message and the Python API.

Why a check could not run

A metadata problem's code can also appear as a check's reason when that problem prevents the check from running. To report an error, include the library version, check ID and a small example that reproduces it.

Choosing an assessor

Assessor(...) raises ProfileError if the requested assessor or version is unavailable. Its code is assessor_not_found or assessor_version_not_found. Use Assessor("FUJI") to select the current default.