Multilingual research dissemination: current practices and implications for bibliometrics

Introduction

Scientific communication is disseminated on a regular basis, both formally and informally, in a variety of formats, to convey knowledge through a lengthy process, from preprint to post‑print as a form of dissemination. This series of activities does not stop when an article is published; rather, it marks the beginning of reaching a broader audience (Garcia et al., 2023). Communication will continue to provide space for exchanging ideas and thoughts, where mutual agreement and understanding of the language used are necessary for effective communication. There is a tacit assumption that English will be the language used in academia (Balula & Leão, 2019). Additionally, the view that English is the language of science because most readers understand it (Dos Santos et al., 2021) and that it has become the most spoken language in the world (Ethnologue, 2025) positions English as a lingua franca in scholarly communication. This enables people to communicate with colleagues from various countries and languages, fostering collaboration. However, it creates a barrier for people who do not speak English.

Effective communication can ensure the flow of ideas and expertise in both directions (Maryl et al., 2020). Using English as the sole language of communication can assist in this with language uniformity. However, it is also important to ensure that other languages continue to play a role in scientific communication, to take into account the language barriers individuals may have. Given that successful dissemination involves reaching an audience with scholarly work that requires the active participation of researchers (Garcia et al., 2023), communicating to a broad audience without regard for language boundaries is key. Dos Santos et al. (2021) highlight the opportunity to enhance the visibility and inclusion of knowledge, while providing assurance to local or regional communities (Dos Santos et al., 2021).

Undoubtedly, research funding can come from resources sourced from the public, which creates a sense of responsibility to help the public understand scholarly work by using language that is accessible to funders. The use of the primary language tends to be more effective for deep understanding, as well as for the creation and sharing of knowledge (Balula & Leão, 2019). Furthermore, publishing in multiple languages not only broadens the audience but also encourages a diverse perspective on research (Kulczycki et al., 2020). Also, the Helsinki Initiative on Multilingualism in Scholarly Communication aims to ensure knowledge is available in various languages to support the dissemination of research results for the benefit of the public (Federation of Finnish Learned Societies et al., 2019). In 2024, G20 recommendations on inclusion, diversity and combating inequalities in science, technology and innovation were announced, which encourage the use of primary languages in science (Directorate‑General for Research and Innovation, 2024).

Providing choices and a variety of languages in scientific communication can facilitate equal access to knowledge for people from all social backgrounds. Using a language understood by the community will be beneficial for the impact and dissemination of the work (Boillos et al., 2024). There are good reasons to preserve linguistic diversity in scientific communication amidst the belief that English should be the lingua franca, such as public engagement and generating impact at the local level (Grange, 2024). Furthermore, publishing research findings in multiple languages can help to develop different perspectives and reach a broader audience (Kulczycki et al., 2020). Additionally, with the rapid advancement of technology and the open science movement, academics can reach a wider audience through the various alternative media available (Garcia et al., 2023). For example, the ability to aid ‘translating’ into different languages utilizing large language models (LLMs), which have outstanding capabilities across many languages (Zhao et al., 2024). However, it is essential to note that LLMs require careful user assessment to evaluate answers, especially in translation, considering that the results may be less accurate (Zhang et al., 2023).

In this article, we compare several platforms that provide opportunities for dissemination by exploring what has already been implemented and how specific rules apply to doing so, if any. While the authors use these platforms, we also look at what several publishers have done to provide options for more than one language in their articles to see if there are any differences in practices and how this affects metrics in general. Finally, we evaluate current practices in accordance with the recommendations made by relevant DOI RAs.

Platforms for research dissemination

Sharing translated research results after an article is published is a complex process influenced by multiple factors. The European Code of Conduct for Research Integrity offers guidance on publishing research in alternative forms, such as translations, to clarify their connection to the original work (ALLEA – All European Academies, 2023). This practice helps readers understand the original article better while maintaining research integrity and avoiding perceptions of improper dissemination. Licensing plays a crucial role in this process. Articles published under a Creative Commons (CC) licence are easier to translate and share due to their flexible terms. In contrast, works under other licensing agreements present additional challenges, as the ability to share translations depends heavily on each publisher’s specific policies. Navigating these considerations is essential to ensure ethical and effective dissemination of translated research.

Currently, there is a wide range of platforms available for research dissemination, each offering distinct advantages, from free platforms to paid services. Academic social networks (ASNs), such as Academia.edu and ResearchGate, are widely used for sharing work with a broader audience. ResearchGate offers free services, while Academia.edu provides a more comprehensive paid version. Another popular option is preprint servers, which function similarly to personal web archives for disseminating work (Gao et al., 2020). Open Science Framework (OSF) Preprints is an open‑source platform that supports many scientific communities, such as SocArXiv, PsyArXiv and LIS Scholarship Archive. Additionally, arXiv explicitly permits the dissemination of translation works, provided they are valuable to the community (arXiv, n.d.‑b). Research can also be shared through various data repositories such as Zenodo, Figshare and Mendeley Data. For example, Chaleplioglou and Koulouris (2023) highlight how Zenodo and OSF Preprints are used as research repositories for ocean and climate science (Chaleplioglou & Koulouris, 2023). A study by Sicilia et al. (2017) found that in Zenodo, text‑based resources, particularly articles, are the dominant type of content shared (Sicilia et al., 2017). These platforms offer diverse methods for disseminating research, with each platform catering to different needs and communities, ensuring that scholars can share their work widely and effectively. Additionally, the DataCite Metadata Properties also provide a mandatory field called ‘ResourceType’ to describe the general type of a resource, with options such as ‘ConferencePaper’ and ‘JournalArticle’ (DataCite Metadata Working Group, 2024). This indicates that certain data repositories support the sharing of more than just research data.

Methods

Given the wide variety of platforms available, this study focuses on examining dissemination opportunities for translated research from the researcher’s perspective, specifically on popular non‑paid platforms. These include ASNs like Academia.edu and ResearchGate, preprint servers such as arXiv and OSF Preprints, and data repositories like Figshare, Mendeley Data and Zenodo. We exclude institutional repositories due to the variability in policies across institutions, although these repositories can also be used. Additionally, we do not consider paid versions of Academia.edu and Figshare+ in this study. To assess the dissemination practices, we reviewed the available guidelines or policies on each platform, where applicable, and uploaded a sample file to explore the fields offered on each platform.

Academia.edu

Academia.edu is a platform designed for researchers to share their work, connect with peers and engage with a global academic community. It allows users to upload and disseminate documents, including both academic and non‑academic content. For this study, we chose to focus on the free version of Academia.edu, as not all users have access to the paid version of the platform. In general, Academia.edu allows users to freely upload various types of documents, and even non‑academic works can be shared. Authors can also utilize multiple languages for their titles, whether by keeping the title in the document’s original language, using translated titles or even combining both. However, there is no specific option to designate a licence for the document. Academia.edu also supports multiple file uploads, allowing readers to access additional versions or formats of the same work. It is important to use distinct file names to differentiate between versions of the document (Figure 1). Additionally, the platform supports the use of multiple files with identical digital object identifiers (DOIs) and provides fields for essential publication details, such as the publication name, publisher, year of publication and abstract, making it a versatile tool for disseminating research.

Figure 1

Zenodo

Zenodo is a data repository that offers a broader range of licences for various resource types, particularly for publications like journal articles and conference proceedings. It also provides publication information for related articles, making it more flexible than repositories like Figshare and Mendeley Data.

Zenodo allows depositors to customize version names, such as labelling a file as a ‘Translated version’, which contrasts with the version number format used by the other two repositories (Figure 2). While versioning is typically used for software updates, Zenodo offers more flexibility in this regard. It also includes a field for specifying the language of the content, supporting multiple language selections and eliminating the need for a version statement.

Figure 2

Unlike the DataCite Metadata Schema 4.5, which allows the language field to appear only once (DataCite Metadata Working Group, 2024), Zenodo enables multiple entries for this field, offering greater customization. This feature aligns with future updates proposed by DataCite to make this field repeatable and recommended (Stathis, 2024). Zenodo can generate DOIs through its system or use the original DOI, and it supports ‘relationType’ to refer to the original source. However, as the Metadata Working Group and DataCite continue to refine the ‘IsTranslationOf’ and ‘HasTranslation’ properties, these repositories do not yet provide a specific option for describing the relationship between the original and translated versions.

Existing practices

The practice of presenting scientific articles in various languages is not new. However, it has not been widespread due to the many factors that may arise in its implementation, such as the cost of publishing being raised (Bachelet & Rousseau‑Portalis, 2023; Sunol & Saturno, 2008), economic inequality for authors in low‑income countries if the responsibility to translate is given to them (Arenas‑Castro et al., 2024) and the consistency of language used (Bachelet & Rousseau‑Portalis, 2023; Pisanski Peterlin, 2019). As Pisanski explained, it is possible to have different translations: those translated by the author and those translated by someone else (Pisanski Peterlin, 2019). We have examined some practices that have been carried out from the publisher’s perspective to provide language diversity in a scientific article. Journals that have implemented this multilingual practice have used different models. Some choose to provide language options for all published articles, while others only offer language options for certain articles, indicating a diversity of rules between them.

The RDBCI: Digital Journal of Library and Information Science adopted a bilingual format (Portuguese and English) in 2016 (RDBCI: Rev. Digit. Bible. and Science. Info., n.d.). Interestingly, RDBCI also includes audio and video content in each published article. Meanwhile, Farmacia Hospitalaria accepts articles in Spanish, has them translated into English and publishes in both languages (ScienceDirect, n.d.). Additionally, BiD, a journal published by the University of Barcelona and Catalonia Open University, while using Catalan as its working language, still accepts articles in other languages. In fact, in its submission guidelines (Section 1.2. Languages), it mentions that manuscripts in certain languages will be published in their original language and a Catalan version will be included if possible (BiD, 2020).

Based on the existing examples, the practice of using diverse languages in publications is not a new phenomenon. However, there are several variations that arise from this practice. RDBCI appears to provide a uniform identifier for all its forms, whether it be the original article, translations, audio or video, all sharing the same DOI, indicating that despite being in various formats, the entries are still treated as a single item. Meanwhile, BiD has different DOIs for the original article and its translation, although they appear as part of a single record. A more distinct practice is observed in Farmacia Hospitalaria, where the original article and its translation are treated as two separate entries, with the note ‘Referred to by’ on the translation version and ‘Refers to’ to indicate the original article.

From the existing practices, it seems that there is no consensus on whether the DOI should remain the same or should differ. What RDBCI does facilitates information retrieval as well as making it easier for readers to access information in various formats with the same identifier, although there may be potential questions about which source was used. However, with different DOIs – as practised by BiD and Farmacia Hospitalaria – it is easier to indicate which version was cited or used. This is useful even though, in practice, individuals generally perform manual translations and still refer to the original article. It is also of use when an article is translated by another publisher; in this case, it is crucial to have different DOIs to identify works from different publishers (even if they are substantively identical), as this will result in different metadata. Another important factor to consider is that both journals must establish a relationship using ‘hasTranslation’ (Feeney, 2023). For example, the paper by Mavili et al. (2015), published in Turkish by Sosyal Politika Çalışmaları Dergisi, was published in English a few years later by Current Psychology, adding ‘Translated article:’ at the beginning of the title (Mavili et al., 2023). This practice resembles what arXiv suggests: that translated articles should have a clear statement in the title.

Implications for metrics

Bibliometrics has become a key tool in academia for evaluating the value of published papers, though it typically requires multiple indicators for a comprehensive assessment. Citation indicators have historically been used in the academic system to measure research impact and influence (e.g. publication counts, citation counts, journal impact factor and h‑index) (Williams, 2022). The h‑index is a widely used scientometric indicator important for several disciplines that work with a simple combination of the number of publications and citations (Kamrani et al., 2021). This indicator is well known for its simplicity, yet significant differences can occur between disciplines, researchers’ age, gender and indexing services such as Scopus and Web of Science (WoS), etc. This becomes problematic because each citation database calculates metrics based on the data they index, rather than considering all citations. Additionally, Bi (2023) highlights four issues with the use of the h‑index: it gives equal credit to all co‑authors, inflating citation counts, creating imbalances in evaluating research contributions and opens the door to unethical authorship practices (Bi, 2023).

This issue also occurs when indexers treat original articles and their translations as separate records, despite having the same content in different languages. For instance, an article in Farmacia Hospitalaria is indexed as two separate entries in Scopus, with one receiving citations while the other does not. The metadata indicates that both have the document type ‘Article’ with the note ‘[Translated article]’ for the translation and identical author keywords and references. Meanwhile, in WoS, both articles have the same title without any additional information to differentiate them. The document type for the translation article is ‘Editorial Material’, and the record for this version does not have any references in its metadata. Even so, other records from this journal show that both the translated and the original have references and the same document type, showing inconsistency.

When an article and its translation are indexed as two separate records, the author ends up with more documents in their profile compared to a publisher who provides only one record for the same article. Citations can affect the author’s h‑index differently depending on which version of the article is cited. For example, if a citation is made to the original article, it will contribute to the h‑index, helping the author reach the required threshold (e.g. an h‑index of three). However, if the citation is directed at the translated version, it will only impact the translation’s record, not the original article’s record. As a result, the author’s h‑index would remain at two, as illustrated in Figure 3.

Figure 3

Another issue arises when the original article and its translation are treated as separate works. Since the records are considered distinct, each version of the article can receive separate citations, meaning the same work could be counted twice for citations instead of once. This creates opportunities for unethical behaviour, as noted by Chapman et al. (2019), who pointed out that such evaluations can be easily manipulated. Researchers might ‘game the system’ by artificially inflating citation counts through this duplication, leading to misleading assessments of their work (Chapman et al., 2019).

Having multiple records for the same work can complicate bibliometric analysis in academia. Since the same keywords are likely to appear in both the original and translated versions of an article, they may be counted multiple times, distorting the analysis. Similarly, when evaluating authors, journals or institutions, extra effort is required to determine whether the records for the original and translated articles should be treated as one. This requires manually checking each record, as relying solely on duplicate DOIs will not work, since the DOI for the original article and its translation may differ.

To accurately assess performance or other aspects of research, additional steps are required to aggregate citations, ensuring they align with the intended metrics. In citation‑based metrics like the impact factor (IF), citation inflation is a significant concern. For example, if an article is cited with a different DOI for its translated version, the journal may receive inflated citation counts (Figure 4), especially when compared to sources that publish only in one language or are multilingual. This highlights the need to reconsider how we assess scientific quality, suggesting that peer reviews of the research’s social impact could be just as important as citation counts (Wallin, 2005). Furthermore, articles published in national languages often cannot be properly evaluated using traditional article‑based metrics (Fuchs, 2014), making it necessary to develop more inclusive measurement approaches, particularly for multilingual sources.

Figure 4

Reflection

Both authors and publishers can provide access to a wider audience through the dissemination of articles in languages other than English. One important consideration for authors is the licensing of their article. In terms of dissemination, Zenodo is the most flexible option compared to others, but there are certainly other factors to consider when making a choice. Clear attribution to the original article is also necessary, either through a title as suggested by arXiv or by providing a note on the translated version and allowing the use of attribution notes such as the related identifiers available in Mendeley Data, Zenodo and Figshare. An overview of multilingual article metadata has been provided by DOI RAs such as Crossref and DataCite. Crossref provides guidance for multilingual registration under a single DOI, whereas separate DOIs are assigned when the content is a translation using the ‘hasTranslation’ relationship (Feeney, 2023). However, Crossref currently states that they still offer limited support for multilingual content (Farley, 2023), which publishers should consider when transitioning to this model. Meanwhile, based on this explanation, the terms ‘Refers to’ and ‘Referred to by’ are deemed unsuitable, indicating that there is still some misunderstanding about the use of relationships in metadata.

Meanwhile, DataCite, one of the providers of DOIs for research data, has proposed a change in which the language can be repeatable and recommended for use, as well as the emergence of new relation types, ‘HasTranslation’ and ‘IsTranslationOf’ (Stathis, 2024), similar to what Crossref offers. For the values of the language, DataCite recommends using IETF BCP 47 and ISO 639‑1 language codes. However, in practice on Zenodo, the values ‘en’ and ‘English’ appear, indicating the need for consistency in whether users are shown in code form (en, es and id) or in full form (English, Spanish and Indonesian). Authors can carry out this practice, so there is no direct control over the metadata because it is dependent on the authors’ understanding, which means that differences and even errors in practice are very likely.

Furthermore, authors must understand that relationship descriptions referencing the original article can be created in the data repository when it becomes available later. However, given each publisher’s different policies, it is necessary to confirm whether the publisher is willing to go through additional steps to provide descriptions in other languages that point to other sources. Given the availability of ‘relationType’, the description of the relationship on both sides (original version and repository) will help readers understand the relationship created. However, it is possible that the description is only on one side, particularly if the publisher’s policy does not address this.

Determining the DOI is also crucial, whether a different DOI is needed or the same as the original article. However, when different publishers or journals are involved, it is essential to have a different DOI, considering that the generated metadata will have an impact. This is in accordance with what Crossref recommends, as it will affect discoverability and citation formatting (Feeney, 2023). Nevertheless, as a means of dissemination, having the same DOI seems important in ASN and data repositories, even though there is an opportunity to generate a new DOI. The lack of agreement or clear guidelines indicates that there are still many paths to take, but caution is needed to avoid falling into the dilemma of self‑plagiarism.

Conclusion

With the goal to reach a wider audience, multilingualism should be taken into consideration by all parties involved in scholarly communication. However, factors such as the quality of writing in different languages, workload, cost and regulations must be considered before proceeding. Publishers can use the approaches described by Crossref to present it as a single item or, if the content is translated, to explain the relationship using relationship metadata. Meanwhile, if publishers have not yet made space for this practice, authors can use various ASNs, preprints and data repositories with an emphasis on complete metadata, such as ‘relationType’ for the primary source and field for language description.

However, in practice, there are still differences between what DOI RAs recommend for publishers and indexers, such as the use of other relationships and metadata inconsistencies. Another issue that may arise is the varying levels of user understanding on this topic, leading to opportunities for inaccuracies. Another factor that could be affected is metrics based on the number of documents and citations, given that there may be differences between the translated scheme and the side that uses multilingualism in its publication. This gap might open up an opportunity for irresponsible actions and raises awareness about the importance of not relying solely on a single metric without considering other factors.

Publishers and indexers have to collaborate to determine the best formulation involving DOI RAs to ensure uniformity of practices in the field, reduce inconsistent actions and clarify relationships from the readers’ perspective. Given that Crossref has previously supported relationship metadata and DataCite’s efforts, DOI RAs must also work together to support dissemination practices in multiple languages and for various platforms to complete their fields according to the recommendations. Support is available, but attention is required to ensure that various parties can implement this without fear of wrongdoing.

CRediT

Faizhal Arif Santosa: conceptualization, data curation, investigation, methodology, visualization, writing – original draft, and writing – review and editing.

Barbara S. Lancho Barrantes: conceptualization, data curation, investigation, methodology and writing – review and editing.

Data accessibility statement

Data supporting this study are published in Repositori Ilmiah Nasional https://hdl.handle.net/20.500.12690/RIN/NS4OKB

Abbreviations and Acronyms

A list of the abbreviations and acronyms used in this and other Insights articles can be accessed here – click on the following URL and then select the ‘full list of industry A&As’ link: http://www.uksg.org/publications#aa.

Competing Interests

The authors have declared no competing interests.