BIBFRAME 2.0 import and export
The bibframe format reads and writes BIBFRAME 2.0 striped RDF/XML. Its resource graph has one top-level bf:Work and one top-level bf:Instance for each entry, linked by bf:hasInstance and bf:instanceOf. A document may contain several such pairs. The importer resolves nested resources and rdf:resource references, as well as typed rdf:Description elements. It does not claim general RDF/XML graph support, such as arbitrary blank-node identifiers or property attributes. Those are alternate RDF/XML syntax forms, not additional bibliographic fields, and are not needed for the pinned converter fixtures. A general RDF parser library was considered; the bounded striped subset and the official reverse converter’s input shape allow this implementation to use the existing XML stack without another runtime dependency. The importer uses StAX to build an in-memory XML tree and resolves the supported resource links from its parsed descriptions.
| JabRef field | BIBFRAME path | Owner |
|---|---|---|
| Title, subtitle | bf:title/bf:Title/bf:mainTitle, bf:subtitle | Instance, then Work |
| Author, editor | bf:contribution/bf:Contribution/bf:agent/rdfs:label with bf:role (aut, edt) | Work |
| Address, publisher, year | bf:provisionActivity/bf:ProvisionActivity/bflc:simplePlace, bflc:simpleAgent, bflc:simpleDate | Instance |
| Edition | bf:editionStatement | Instance |
| Series | Work bf:relation with series relationship and bf:Series/bf:title, or bf:seriesStatement | Work relation or Instance |
| Page count | bf:extent/bf:Extent/rdfs:label with a page unit | Instance |
| ISBN, DOI | bf:identifiedBy/bf:Isbn or bf:Doi/rdf:value | Instance |
| ISSN | bf:identifiedBy/bf:Issn/rdf:value | Instance or journal host Work |
| Language, abstract | bf:language, bf:summary/bf:Summary/rdfs:label | Work |
| URL | bf:electronicLocator or bf:supplementaryContent/bf:SupplementaryContent/bf:electronicLocator | Instance, then Work |
| Linked online file | bf:hasInstance/bf:Instance/bf:electronicLocator on a secondary Instance | Work secondary Instance |
| Journal, pages | bf:relation with partof host Work typed bf:Serial; host Instance bf:part | Work relation |
The pages: prefix in bf:part follows the pinned converter’s article output. Other BIBFRAME readers may not interpret it as a page range.
A Work with a serial host becomes an Article; the host is identified by bf:Serial type or an ISSN. Other linked Works become Books. A host title alone is not enough to identify a journal. Recognized single-language names and codes export as MARC language URIs; other language text exports as a literal label. Data outside this table, such as MARC administrative fields, physical description other than page count, subject headings, and Instances other than linked online files, is lost. Citation keys are never inferred from resource URIs. The exporter hashes the entry type and mapped exported fields with SHA-256 to mint stable Work and Instance URIs. Citation keys are excluded. Identical entries in one export receive numbered suffixes so their resources remain separate. ISBN, ISSN, and DOI values are preserved as bibliographic identifiers. HTTP(S) linked files use secondary Instances. Their titles become file descriptions; an absent title leaves the description empty. A .pdf locator identifies a PDF link; otherwise the file type remains unspecified. Local file paths stay in JabRef; they are not exported as BIBFRAME locators.
The fixture commands and pinned official converter commits are recorded beside the fixtures in jablib/src/test/resources/org/jabref/logic/importer/bibframe/README.md. The reverse converter accepts one description per invocation, so split a multi-entry export into linked Work/Instance pairs before passing it to that converter.
The pinned reverse converter emits an article host as MARC 773 without subfield 7. The MARC parser can still recognize a serial host when 773$x contains an ISSN; without either signal, the article type and journal are lost at the final MARC import stage. It does not infer Book from the MARC leader either. Language is emitted only in MARC 008, which that parser does not read. The MARC parser removes the converter’s trailing period from publication dates. Four-digit years also export as structured EDTF bf:date, which the reverse converter uses for MARC 008 date positions. A generic URL becomes MARC 856 without the Volltext label that the parser requires for a linked file. The reverse converter may add ISBD punctuation to title and publication fields.