Multilingual Metadata Is Now a Journal Visibility Workflow

A bilingual journal can serve its readers well and still confuse the systems that make scholarship findable. The article page may show the title in Portuguese and English. The PDF may carry a Spanish abstract. The DOI record may contain one language because the export path cannot express both cleanly. The index may display whichever string arrived first. Nobody set out to erase a language. The workflow simply chose one field where the scholarship needed a relationship.

That is why the 2026 metadata conversation around Latin America deserves attention outside the region. Crossref reported that its second Metadata Sprint took place in Sao Paulo from March 4 to 6, 2026, with 31 participants from Argentina, Brazil, Colombia, Ecuador, and Mexico. Crossref also described it as its first tri-lingual sprint, with people moving between Spanish, Portuguese, and English while working with open scholarly metadata.

This was not a symbolic language-access event. It was a practical signal about how scholarly infrastructure has to behave when research communities are multilingual by default. If a journal publishes in more than one language, translates article metadata, or serves authors whose names and institutions do not fit English-language assumptions, multilingual metadata is no longer a nice enhancement. It is part of visibility, integrity, and assessment.

The Sao Paulo Signal

Crossref announced the sprint as its first in Latin America with SciELO participation and said the event would be held in Portuguese, Spanish, and English to support active participation. After the sprint, Crossref highlighted the role of SciELO as a major regional source of open scholarly content and metadata across articles, books, preprints, and datasets published in different languages.

For journal leaders, the important point is not that a workshop happened. The important point is what the workshop made visible. Multilingual publishing is not an edge case at the margins of the global record. It is one of the conditions under which the global record is produced. Editors, librarians, developers, researchers, and scholarly communications teams were not just translating instructions. They were asking how open metadata can represent real editorial practice.

That distinction matters. Translation work often happens late: after acceptance, after copyediting, after the article page is almost ready. Metadata work has to happen earlier. The system needs to know whether a second title is a translation, a parallel title, an alternative title, or a separate work. It needs to know which abstract should be displayed where, which language the article itself uses, and whether a translated version requires its own persistent identifier or a relationship to the original record.

Where Language Gets Flattened

The weak point is usually not the public page. Journal teams often care deeply about showing the right language to readers. The weakness appears when the record leaves the page and enters metadata deposits, indexing feeds, APIs, repository exports, dashboards, and discovery services.

Titles and abstracts become single-choice fields

DOAJ states in its metadata FAQ that its XML format currently supports only one language for article title and abstract, while it researches a solution for multiple languages. It also warns that metadata containing multiple languages can be uploaded, but publishers cannot choose which language is displayed and should send one language to avoid mixed-language displays. That is a perfectly reasonable operational constraint. It is also a reminder that journal workflows cannot assume every downstream system will preserve the language choices visible on the journal site.

Names and institutions lose local precision

Author names, department names, institutional names, funder names, and place names carry local meaning. If staff normalize them too aggressively for an English-language export, they may make matching easier in one system while weakening the record in another. If they leave every variant unmanaged, the same author or institution may fragment across search and assessment tools. The answer is not to pick purity over interoperability. The answer is to record the local form, the preferred display form, and the identifiers or relationships that make matching less fragile.

Translations are treated as decoration

Crossref metadata can express relationships such as "is translation of", and its March 2026 metadata-enrichment discussion lists translations among the relationships that help form the research nexus. Many journal workflows still treat translated titles and abstracts as page content rather than structured relationships. That leaves discovery systems to infer whether two records are connected, duplicate, or unrelated.

Discovery Equity Depends On Export Discipline

DOAJ opened its 2026-2028 strategy by saying it wants quality open access scholarship from all regions, languages, and publishing traditions to be visible, discoverable, and valued. Its goals include broader integration of DOAJ metadata into discovery, assessment, and research infrastructure systems, plus multilingual guidance and support for smaller and emerging publishers.

Those ambitions will not be met by good intentions alone. They depend on export discipline: title language, abstract language, journal language policy, article language, contributor identifiers, affiliation identifiers, license metadata, references, and translation relationships need to survive handoffs. A regional journal can publish excellent bilingual scholarship and still be disadvantaged if its English title travels to one service, its local-language abstract travels to another, and neither record carries the relationship that explains the pair.

This is also where assessment risk enters. Discovery services, analytics vendors, library systems, AI search tools, funder dashboards, and institutional repositories increasingly reuse open metadata. If the multilingual record is thin, those systems may undercount the journal, misclassify the language of publication, split usage and citation context, or make local scholarship look less connected than it is. The damage is subtle because the article remains online. The loss happens in the graph around it.

A Bilingual Article Needs One Governed Story

The right operational question is not "Which language should we send?" It is "What story should every downstream system be able to reconstruct?" For some journals, the article has one primary language and translated metadata for discovery. For others, the full text exists in two languages. Some publish translated versions as separate citable objects. Others use parallel title and abstract fields to serve local and international readers from one article page.

Each model can be valid. The problem is switching models article by article without recording the difference. If a translated abstract is editorially reviewed, say so in the workflow. If a translated title is for discovery only, store that distinction. If a full translation has its own DOI, link the records. If the journal sometimes publishes in English, sometimes Spanish, and sometimes both, capture article language as structured data instead of inferring it from the PDF.

Editors do not need to become schema specialists, but they do need a local rulebook. Without one, production staff make case-by-case decisions under deadline, and those decisions become the permanent metadata record.

The Audit To Run Before The Next Issue

• Pick ten recently published articles that include more than one language in the title, abstract, keywords, article page, or full text.
• Compare what appears on the journal page, in the PDF, in the DOI metadata, in DOAJ or another index, and in the repository feed.
• Record which language is treated as primary, which translations are present, and whether language tags or translation relationships are explicit.
• Check whether author and affiliation names preserve local forms while also carrying ORCID and ROR identifiers where available.
• Identify one downstream system where the display is wrong, mixed, or incomplete, and trace whether the issue began in author submission, editorial metadata capture, production export, or the receiving platform.
• Write a short rule for future articles: how titles, abstracts, keywords, translated versions, and article language should be captured before publication.

The point of the audit is not to shame the platform or the index. It is to find the places where editorial reality becomes ambiguous data. Once those places are visible, the journal can fix collection forms, staff instructions, export mappings, and exception handling.

A Leadership Move, Not A Translation Task

Multilingual metadata sits awkwardly between editorial policy, production, technology, and community mission. That is exactly why it needs leadership attention. A managing editor can correct a title. A production vendor can adjust an export. A metadata librarian can diagnose a bad record. But only leadership can decide the standard the journal owes its authors and readers across languages.

The practical takeaway for journal leaders is simple: choose one multilingual publishing model for the journal, write down how it should appear in every public and machine-readable surface, and test the next issue against that model before publication. If the local-language record and the international discovery record tell different stories, the journal has not translated its scholarship. It has fragmented it.