Navigation – Plan du site

AccueilProjets collaboratifsCybergeoNet. Scientific translations

CybergeoNet. Scientific translations

Christine Kosmopoulos, Natacha Aveline, Colette Cauvin-Reymond, Bernard Elissalde, Maria Gravari-Barbas, Margaux Hardy, Nathalie Lemarchand, Francisco Maturana, Denise Pumain, Benjamin Wayens et Gloria Zamorano
Cet article est une traduction de :
CybergeoNet – Traductions scientifiques
Autre(s) traduction(s) de cet article :
CybergeoNet. Traducciones científicas

Notes de la rédaction

This article is the translation of: CybergeoNet. Traductions scientifiques

Texte intégral

Project background: Cybergeo and translation

1When Cybergeo, the European Journal of Geography, was created in 1996, it immediately adopted a multilingual interface and has since published in several European languages. Indeed, the journal's ambition is to have a strong international visibility, its creation wished for by the informal network that has been holding European colloquia on Theoretical and Quantitative Geography every two years since 1978 (150 to 200 participating academics). Today, the journal receives proposals for articles from all over the world and publishes about fifty articles per year (of which about a third are by foreign authors).

  • 1 Brunet, R., Ferras, R., & Théry, H. (1993). Les mots de la géographie : dictionnaire critique. Recl (...)

2Specialized in geography, Cybergeo covers all the branches of this discipline whose lexical field is quite specific among the human and social sciences: in addition to concepts developed by the social sciences, it includes vocabularies of physical geography, environment and planning, specific statistical methods (auto-correlation, fractals, spatial analysis), very varied sources such as satellite images, massive geo-referenced data, own instruments (cartography, geographic information systems), headings intended to promote a reproducible science (data papers and model papers). Several dictionaries mention this richness and sometimes a certain variability of definitions1.

3This specificity within the humanities and social sciences makes the issue of translation particularly complex and costly for a journal specialized in geography. With its international ambition, the journal has nonetheless tried to meet this challenge for several years. It has pragmatically developed an in-house method to ensure its translation needs, as it does not have sufficient means to pay professional translators. Managed with the resources of the Géographie-cités laboratory, the journal now functions with one full-time person (a research engineer who performs the two functions of editor-in-chief and editorial secretary paid by the CNRS) and an editorial team whose volunteer members come from the laboratory and an editorial board composed of scientists from other universities and research institutions (Who are we?).

4Here is a brief reminder of the different steps that led to the current functioning of the translation:

  • Plurilingual from the beginning in 1996, with an interface in French and English, titles, abstracts and keywords in French and English and in another European language of publication of the article (Spanish, Italian, German, Greek, Hungarian, Swedish...).

  • Translation into English (since the beginning) and Spanish (since 2014) of all editorials.

  • By joining the Revues.org portal in 2007, Cybergeo brought its multilingual model for Lodel developed by OpenEdition.

  • In 2014, a Spanish interface was added to French and English.

  • In 2018, Chinese translation of metadata (titles and abstracts of recent articles) is launched.

  • Since 2011, a program of translation of selected articles into English and Spanish (29 articles translated in 7 years) has been funded through membership in the Freemium model. The translations were carried out by volunteer researchers and temporary staff.

  • Faced with the insufficient quality of foreign language abstracts and keywords produced by authors for their articles, the journal began a retrospective program in 2016 to check and correct English abstracts and keywords for its approximately 880 published articles.

5All these investments in internationalization were made possible largely by volunteer involvement and for the rest by funding received with Freemium since 2011 (i.e., about 6000 euros/year). The journal also receives a subsidy of 3000 euros/year from the CNRS for its operating costs. The current budget does not allow for the payment of professional translators.

Description of the project

The project

6CybergeoNet is a laureate of the AAP "Scientific Translations" launched in 2019 by the Ministry of Higher Education, Research and Innovation (MESRI) in France. It is carried by Cybergeo: European Journal of Geography in partnership with two international journals Brussels Studies and ViaTourism and the digital encyclopedia of geography Hypergeo.

7Starting from the specialized vocabulary of geography, CybergeoNet has developed a protocol for writing and translating metadata that can be adapted to other disciplines in the humanities and social sciences. The aim was to systematize actions already tested in terms of translation and to test methodological advances shared with other journals in order to access, through translation, a better international visibility of their publications without any additional cost for journals with small budgets like Cybergeo.

Actors/team

8Project manager
Christine Kosmopoulos, Research engineer at CNRS, editorial manager of Cybergeo

9Participants

  • Natacha Aveline, Research Director at CNRS

  • Colette Cauvin-Reymond, retired university professor

  • Bernard Elissalde, Professor emeritus at the University of Rouen and director of Hypergeo

  • Maria Gravari-Barbas, Professor at the University of Paris 1 Sorbonne and director of ViaTourism

  • Margaux Hardy, Researcher, Université Saint-Louis - Brussels and editor of Brussels Studies, Belgium

  • Nathalie Lemarchand, Professor at the University of Paris 8 and vice-president of the IGU

  • Francisco Maturana, Professor at the Universidad Austral de Chile, Chile

  • Denise Pumain, Professor Emeritus at the University of Paris 1 Sorbonne and Scientific Director of Cybergeo

  • Benjamin Wayens, Professor at the Université Libre de Bruxelles and Associate Editor of Brussels Studies, Belgium

  • Gloria Zamorano, Professor at the University of Mendoza, Argentina

  • Liubing Xie, Chinese translator

Languages and corpus type (metadata, abstract, full text)

10The tests focused on the translation of metadata into English, Spanish and Chinese for Cybergeo, Hypergeo and ViaTourism articles. Full text translation tests were also undertaken in Spanish and English for Cybergeo articles, and in Arabic and Chinese for ViaTourism.

Tools and software used

11The tests with English and Spanish compare human translation and machine translation of metadata. The machine translation tool mainly used is the partially free software DeepL, but in some cases Google Translation proved to be an equally useful and complementary tool.

List of documents produced

  • Final report on the CybergeoNet project (64 pages) with its annexes (323 pages) including:

  • Cartographic glossary - Colette Cauvin-Reymond

  • Multilingual tourism glossary from ViaTourism

  • Best practices guide for writing and translating metadata

  • Cybergeo multilingual index

  • Work and organization protocol for editorial teams

  • Impact report

Localization

12All the deliverables (report and annexes) can be downloaded on HAL : https://halshs.archives-ouvertes.fr/​halshs-03419355 and the data in Cybergeo Harvard Dataverse Collection : https://dataverse.harvard.edu/​dataset.xhtml?persistentId=doi:10.7910/​DVN/​KGZTIT

License to use

13All documents and data are available under the creative commons 4.0 international license.

Progress of the project stages

14The first working meeting in early December 2019 brought together all the partners. It was an opportunity to structure the organization by working group according to the five axes of the project. During 2020, the working groups met monthly to discuss the progress of the work. At the end of 2020, a mid-term report was submitted. During 2021, an impact study was launched on the improvement of the quality of metadata and the translation of full-text articles, which was accompanied by the writing of the final report. In November 2021, the detailed project report and its annexes were finalized. A 6-page summary report was also written.

The project is divided into five work areas

  • Improving the quality of metadata translations of articles already published in Cybergeo by having existing translations checked and corrected by specialists and by occasionally using professional translators;

  • Evaluate the contribution of machine translation methods for scientific translation;

  • Develop from this experience an innovative, efficient and collaborative translation method that can be transferred to other journals with a similar economic model (autonomous, open access and with a small budget);

  • To ensure that authors and editorial teams become more competent in translation and metadata writing;

  • To test the impact of the improvements on the international visibility of the journal through OpenEdition's consultation statistics.

Methodology and results

On the improvement of metadata (axis 1)

15A young non-French-speaking Chinese specialist of English language and literature, Liubing Xie, was asked to translate into Chinese the metadata (title and abstract) of the articles published in Cybergeo from the English version. Natacha Aveline took care of the Chinese proofreading in order to guarantee a faithful reproduction of the scientific field. Independently of her knowledge of Chinese, she also used Google translation to compare the different French/English/Chinese translations. The quality of the translation from Chinese into French was quite satisfactory due to the proximity of the syntactic structure of Chinese to that of French and English.

16The translation of the metadata by Liubing Xie appeared to be of a very good level overall, but some difficulties were encountered. Some of the problems stemmed from the original English text, which was often too vague or jargonous, leading to confusion. Liubing Xie had restructured several mis-stated sentences in the original text during the translation process, and in some cases N. Aveline relied on her Chinese translation to modify the English abstract! Liubing Xie also indicated that, although he does not speak French, he sometimes used the French text with the help of Google Translate, to try to understand what the author had meant in the English version. However, we could see that the French summaries were also problematic, so we had to go back to that level to check the translation chain.

17Results: they then decided to proceed in the opposite way of the initial approach. The French and English abstracts (author's version) were first sent to N. Aveline for verification, if necessary with the help of DeepL for the English version, knowing that we also had a final proofreading by a native English speaker. The English version of the abstract was then sent to Liubing Xie.

On the contribution of machine translation methods to scientific translation (axis 2)

18The verification of the metadata translated into English led to the observation of structural problems that may concern the size of the abstracts (often too long), the poor readability of the problematic and the results, the not very explicit character of the method, and/or a not very relevant choice of keywords written by the authors of the published articles. In such cases, it was necessary to rework the French version of the abstract and sometimes the keywords.

19This version was then submitted to machine translation (MT) in English by DeepL, to which possible adjustments could be made. It should be noted that the final proofreading by our English-speaking proofreader did not add any significant value to the translation of DeepL under our control, attesting to the high level reached today by these automatic translators between French and English.

20Problems of wording/style could be detected depending on the language. Although the French formulations, more humble, are quite correct, they are not very compatible with the economy of words required by the format of the abstracts and can give an impression of hesitation to a readership acculturated to international standards. The choice was therefore made to bring the abstracts into line with Anglo-Saxon standards for metadata. It should be pointed out that automatic translators do not intervene in these differences in wording.

21The studies conducted by C. Cauvin on the metadata of articles compare DeepL to the translation done at the time of publication of the article by a translator. Starting from the English translation, the summary was re-translated into French with DeepL, and then in both directions from the initial French version with DeepL. Confusions of a semantic nature on a term (power) were observed, which prompted a return to French after the English translation to avoid involuntary errors.

22F. Maturana and G. Zamorano used the same method as described above for English. They used Google Translation (GT) and DeepL for the Spanish translation of the metadata of the published articles. Specifically, they entered the French text and automatically translated it on Google Translation and DeepL. Overall, few problems were encountered; however, it was often necessary to modify inappropriate words or to repeat some expressions. There were also choices to be made between suggestions from GT and DeepL. In all cases, proofreading of the translation by a specialist in the language and field is essential to ensure an accurate translation. In some cases, these translators, geographers in Spanish, in this case, call upon colleagues when very specific terms are involved (on satellite images, physical geography, etc.).

23The importance of a clear, precise and well-structured original text for a good translation is emphasized here again. Before undertaking translations of summaries, it is important to note things that are not clear in the original language. It is at the time of translation that confusions or even contradictions in the original text, abstract or article, become apparent.

24N. Lemarchand and B. Elissalde based their observations on the translation of full texts of Cybergeo and Hypergeo and compared the English and Spanish translations in both directions. Because the algorithms work according to a probability of use calculation, MT texts (DeepL and GT) tend to choose the most frequent lexical correspondences. Machine translation thus tends to produce normative, which can be interesting for the recording of metadata but impoverishing for the authors' writing. The classic bias of MT is to move towards the greatest simplicity of the language, which has certain advantages but could also lead us to ask ourselves whether we should not invent a "Google scholar" of MT? To date, the tests clearly show the risk of impoverishing the language through automatic translation with a simplified vocabulary and syntax, even too simplistic for entire articles.

25Results: in conclusion, the high quality achieved by automatic translation software allows metadata to be rendered satisfactorily in French, English and Spanish, provided that human proofreading is carried out before the articles are put online. Efforts must also be made to ensure the quality of the original title and abstract in order to avoid structural problems, wording errors and the use of false friends. On the other hand, in order to preserve the richness of the language, the style and the argumentative construction, it is essential that the authors take care to check the translation of their full text.

On the innovative translation method (axis 3)

26CybergeoNet is committed to providing an innovative, efficient, collaborative and transferable translation method to other journals with a similar economic model, in open access with small budgets.

27The protocol has been developed including a working method for the translation of metadata on the fly for journal editorial teams to ensure a good quality of translation at low cost. The protocol is based on the editorial practices implemented at Cybergeo, described below, on Cybergeo's multilingual keyword index and on glossaries specialized in cartography and tourism.

28In the note to authors published on the Cybergeo journal website, a best practice guide on metadata writing and translation has been inserted. This guide is sent by the editorial teams to all authors whose article has been accepted for publication in order to help them to conform the metadata of their article, the writing and the translation in the different languages requested.

29When the article corrected according to the recommendations of the editorial board and the best practice guide on metadata editing and translation is returned by the authors, it is sent to the editor of the journal. The verification of the quality of the writing (in the original language) and the translation of the metadata into English, Spanish and, as the case may be, Chinese is done by several geographer experts and an English-speaking translator.

30At this stage, the preprint can be sent back to the authors by the editorial service with a request for improvement of the metadata, the title and/or abstract and/or the keywords. The authors will have provided, at the request of the journal, specific terms on the theme of their article with proposals for translations into English and Spanish. In the case of Cybergeo, the journal only adds a keyword in very exceptional cases. Indeed, even if the content of science is continuously renewed, the software of the current OpenEdition platform does not allow to modify the list of keywords by applying an automatic retroactive update. Only completely new concepts are considered for integration. Currently, Cybergeo's multilingual index contains nearly 2200 keywords describing its articles.

31Results: The multilingual index of Cybergeo's keywords and the glossaries specialized in cartography and tourism are available in open access on HAL. They can be enriched with the list of specific accepted terms provided by the authors and will be shared between the editorial teams.

On the rise in competence of the authors and the editorial teams in the field of translation and writing of the metadata (axis 4)

32In the perspective of ensuring a better international visibility to their authors, CybergeoNet journals share the ambition to provide quality metadata, which the authors of articles write most of the time in an empirical way without any particular instructions.

33The first meeting of CybergeoNet in December 2019 highlighted the need to form a working group to reflect on recommendations for writing metadata (title, abstract, and keywords) that play a decisive role in the dissemination and impact of the article. The working group, composed of members of the three journals (Cybergeo, Brussels Studies and ViaTourism), was set up in January 2020 with the objective of drafting a best practice guide for authors and editorial teams. Several requirements have been retained: a clear structure and language highlighting the salient points of the article, a syntax and vocabulary that can be easily analyzed by search engines without impoverishing the meaning, and finally a scientifically sound translation.

34Since the three types of metadata (title, abstract and keywords) each have a specific function, it was agreed that the choice of words and writing should be specific to each category, while recognizing that they are complementary. Indeed, the insertion of an important word in the content of the title and the summaries has as much impact as in the list of keywords, which in principle is closed. The specific word(s) of the article can be introduced in the title and/or the abstract, which will also be analyzed by the search engines, for example for rare place names that describe very few articles in the whole journal.

35The aim of the small guide to good practice in writing and translating metadata is to help authors adopt a rigorous writing style with concepts that can be easily analysed for retrieval and access to metadata on the web, without however proposing forms and structures that would be too stereotyped and without excessively reducing the richness and diversity of the language

36Results: the recommendations in this guide are incorporated into the note to authors posted on the journals' websites. When an article is accepted for publication, authors are again asked to rigorously apply the advice of the good practice guide that will be sent to them.

On the impact of the improvements on translation (axis 5)

37The 16/2019 issue of ViaTourism has included for the first time translations into Chinese and Arabic. Publishing the articles in these two new languages allowed us to observe through the statistics if the readership is amplified in the Arab countries and China. In general, a comparative examination of the statistics between the years 2019 and 2020 (from September to September) reveals a significant increase in visits to the site from China with a notable rebound in visits. Over the period between April and September 2020, the total number of visits to the journal hosted on OpenEdition increases from 499 in 2019 to 4,270 in 2020.

38Colette Cauvin has produced a detailed report based on statistics provided by OpenEdition on the impact of metadata translation and certain articles on Cybergeo visits, which leads to several findings. Overall, between 2016 and 2020, the site's traffic, starting from a high level in 2016 (the year of the journal's 20th anniversary celebration), dropped sharply in 2018 to rise sharply again in 2020, but this recovery is not necessarily linked to translations. The spatial approach, more adapted to highlight the impact of translations, has allowed us to identify countries interested in the journal, which have progressively increased their consultation. Two classifications, one quantitative, the other linked to the rhythm of consultations, revealed the fundamental role of English, which attracts both English-speaking countries (India, Southeast Asia, Australia, etc.) and European countries (Germany, Czech Republic, Poland). Translations into Chinese seem to have had a real impact in China, a country that is ranked 5th in 2020. The Spanish metadata concern not only Spain but also Latin American countries (Argentina, Colombia) or Central American countries (Mexico). However, these conclusions must be qualified because the translations are still recent and it would be necessary to extend this study to find out whether in the coming years concerned countries remain faithful and whether other countries are emerging.

Conclusions and perspectives

39Several interesting results emerge from the pooling of translation problems encountered by low-budget geography journals. The most important result is that the translation operation still requires human control, precisely because of the specificity of scientific vocabulary. The risk of error is renewed with the emergence of new concepts or expressions diverted from their common usage to build and refine the notions worked on in the humanities.

40That said, a second equally important result underlines the relative reliability of machine translations, which save precious time for volunteer translators. Provided that a final proofreading is carried out, the translation operation is greatly simplified by the use of DeepL or Google translation in particular. The process is validated, from French to English or Spanish, but also to Chinese. An interesting initiative is to point out the interest of a back-translation in the original language in order to check that no misunderstandings have been introduced.

41A third interesting and unexpected result is the attention that the translation exercise brings to the quality of the original metadata, by forcing to purge the titles and abstracts of their inconsistencies, imprecisions or forms of writing that would be sources of ambiguity for the translation. The quality of scientific communication in an international environment is greatly improved by this awareness of the importance of metadata for search engines. On the other hand, the alignment that it implies with Anglo-Saxon editorial standards carries the risk of a certain impoverishment of the subtleties of writing that can be formed in other languages, including English!

42Another interesting and longer-term result of this exercise is in the rapprochement between practices and the inter-knowledge that results from collaboration between French-speaking journals. The discussion of problems and the sharing of solutions, in the form of best practice guides and multilingual and evolving interactive glossaries are results that show the fruitfulness of a cooperative approach to scientific publishing.

Haut de page

Notes

1 Brunet, R., Ferras, R., & Théry, H. (1993). Les mots de la géographie : dictionnaire critique. Reclus La Documentation française. Lévy, J., & Lussault, M. (2013). Dictionnaire de géographie et de l'espace des sociétés. Belin. Marei, N., & Richard, Y. (2018). Dictionnaire de la régionalisation du monde. Atlande. Pumain D., Paquot T. & Kleinschmager R. (2006), Dictionnaire La ville et l’urbain, Anthropos.

Haut de page

Pour citer cet article

Référence électronique

Christine Kosmopoulos, Natacha Aveline, Colette Cauvin-Reymond, Bernard Elissalde, Maria Gravari-Barbas, Margaux Hardy, Nathalie Lemarchand, Francisco Maturana, Denise Pumain, Benjamin Wayens et Gloria Zamorano, « CybergeoNet. Scientific translations », Cybergeo: European Journal of Geography [En ligne], mis en ligne le 04 mars 2022, consulté le 29 mai 2023. URL : http://journals.openedition.org/cybergeo/38400

Haut de page

Auteurs

Christine Kosmopoulos

Inégnieure de recherche, CNRS-UMR Géographie-cités, France

Articles du même auteur

Natacha Aveline

Directrice de recherche au CNRS, CNRS-UMR Géographie-cités, France

Articles du même auteur

Colette Cauvin-Reymond

Professeure d’université à la retraite, France

Bernard Elissalde

Directeur d’Hypergeo, France

Articles du même auteur

Maria Gravari-Barbas

Professeure à l’Université Paris 1 Sorbonne et directrice de ViaTourism, France

Articles du même auteur

Margaux Hardy

Chercheuse, Université Saint-Louis – Bruxelles et éditrice de Brussels Studies, Belgium

Nathalie Lemarchand

Professeure à l’Université de Paris 8 et vice-présidente de l’UGI, France

Francisco Maturana

Professeur à l’Université de Santiago, Chili

Denise Pumain

Professeure émérite à l’Université Paris 1 Sorbonne et directrice scientifique de Cybergeo, France

Articles du même auteur

Benjamin Wayens

Professeur à l’Université Libre de Bruxelles et rédacteur en chef adjoint de Brussels Studies, Belgium

Gloria Zamorano

Professeure à l’Université de Mendoza, Argentina

Articles du même auteur

Haut de page

Droits d’auteur

CC-BY-4.0

Creative Commons - Attribution 4.0 International - CC BY 4.0

https://creativecommons.org/licenses/by/4.0/

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search