Publication, Languages, and the Reuse of Knowledge

A scholarly text exists together with its source, edition, language, terminology, bibliographic description, structure, rights of use, and history of changes. These details determine whether a document can be found, whether a specific fragment can be cited, whether it can be compared with another version, corrected, translated, or incorporated into a new corpus.

When editorial decisions remain only in the final PDF, part of the work has to be repeated at every later use. Textual infrastructure preserves those decisions as separate, described, and versioned resources.

Producing Knowledge Requires Infrastructure

The global research system has substantial resources, distributed unevenly across countries and institutions. According to the U.S. National Center for Science and Engineering Statistics, global research and development expenditure in 2024 was about 3.48 trillion dollars at purchasing power parity. China accounted for 30% of that expenditure, the United States for 29%, and the European Union for 17.6%.¹ In the same year, the observed corpus included about 3.5 million scientific and engineering articles; 31% were attributed to China, 12% to the United States, and 7% to India.²

These indicators describe the productive capacity of research systems. They do not measure the quality of a particular article and do not allocate intellectual value by country. For textual infrastructure, other consequences matter: large systems can maintain publishers, repositories, bibliographic services, terminology databases, translation units, software platforms, and long correction cycles.

Data from the UNESCO Institute for Statistics show another level of disparity. In 2023, Europe and North America had an average of 4,358 researchers in full-time equivalent per million inhabitants, while Sub-Saharan Africa had 88. Seventy-three percent of countries spent less than 1% of GDP on research and development.³ These figures create different capacities for long-term support of specialized literature, even when the initial publication is open.

The Language of Scholarly Work

English serves as the common working language for much of international science. It reduces coordination costs among researchers, journals, and conferences. At the same time, it shifts additional work onto authors and readers for whom English is not a first language.

The study by Tatsuya Amano and co-authors covered 908 researchers in ecology, evolutionary biology, conservation, and related disciplines from eight countries.⁴ Among participants who had published one English-language article, researchers from countries with medium English proficiency spent a median 46.6% more time reading an article, and those from low-proficiency countries 90.8% more time, than native English speakers. For preparing the first draft of a paper, the corresponding estimates were 50.6% and 29.8% additional time.⁵

In the same sample, 38.1% of researchers from medium-proficiency countries and 35.9% from low-proficiency countries reported having experienced rejection specifically because of the quality of English writing; among native English speakers, that experience was reported by 14.4%. Frequent requests to improve English during peer review were reported by 42.5% and 42.6% of non-native speakers, against 3.4% of native speakers.⁶

The study concerns specific disciplines, countries, and participant self-reports. Its results cannot be transferred to all researchers or every field. It does, however, clearly show the mechanism: international language infrastructure distributes time and editorial costs unevenly.

Open Publication and Reuse

Open access answers the question of lawful and free access to a publication. Further work with the material requires additional properties: a stable address, usable metadata, a clear licence, citable structure, provenance information, and formats that can be processed outside the original platform.

UNESCO’s Recommendation on Open Science defines scientific knowledge as openly accessible, usable, and reusable. Necessary components include repositories, open publishing platforms, identifiers, standardized services, and interoperable infrastructures.⁷ The FAIR principles state the same task at the level of data: an object must be findable, accessible through a standard protocol, linked to formal vocabularies, and supplied with a licence and provenance.⁸

For a textual project, these requirements become concrete editorial objects:

• bibliographic record for the source;
• stable identification of the work and fragment;
• text version and correction log;
• separate attribution of author, translator, and editor;
• links among original, translations, and annotations;
• rules for transmitting variants;
• machine-readable export;
• conditions for reuse.

A public file becomes a durable resource once those details can be extracted and transferred with the text.

Multilingual Editorial Work

Each language version creates a recurring set of operations. The editor determines segmentation, selects terminology, checks names and citations, links parallel fragments, records divergences, and updates the material after a source correction. When processes are not coordinated, these actions are repeated separately for each language.

The European Union has 24 official languages. Directly connecting each language with every other gives 552 translation directions; the European Parliament also uses relay languages and centralized infrastructure.⁹ Translation memories, terminology databases, and parallel corpora make it possible to reuse already verified segments. The European Commission’s eTranslation service was trained on decades of institutional translations and accepts structured formats, including TMX and XLIFF.¹⁰

The scale of the Réseau-Observatoire is far smaller than that of the European Union institutions. For a small project, it is especially important to preserve work already done: recreating it is usually more costly than maintaining a properly structured registry.

Reusable Textual Representations

A textual version can serve both as reading material and as a structured layer of a corpus. To do so, it is connected with the source, segmented, supplied with stable terminology, and accompanied by information on editorial responsibility.

source and bibliographic record
                 ↓
stable segmentation and terminology choices
                 ↓
translations, annotations, and normalized representations
                 ↓
parallel corpus · terminology database · editorial history
                 ↓
new editions · dictionaries · research · language tools

Such an architecture does not eliminate the intellectual work of the translator and editor. It makes the results of that work visible and reusable.

Romance Branch of the Program

The Réseau-Observatoire maintains a normalized lexico-syntactic representation on a Romance basis. It is applied in selected reading versions, terminology records, and parallel texts. The profile sets constraints on lexical bases, productive derivation, neutral component order, and the expression of selected grammatical relations.

The complete textual implementation of the active revision is used in language and publishing resources under the name romanova. The profile specification, general language documentation, and corpus of applications are published separately. This allows a thematic project to cite a specific normalized version without turning its own page into a language description.

The work of the Réseau-Observatoire consists in accumulating described objects: standards and profiles, texts and terminology records, parallel versions and editorial provenance. Their value increases as one prepared resource becomes the source material for the next.