Since 2011, OpenEdition Center has been developing the OpenEdition Lab research and development programme, which aims to develop and test a range of tools exploiting the possibilities of digital in terms of collaboration, access to scientific information, and information retrieval in different repositories such as books, journal articles, research blog posts and scientific programmes.

The objective is to enable the deployment of innovative services through new mechanisms facilitating the dissemination of scientific information as well as the implementation of research and development projects aiming to link documentary resources together.

Research and development projects (2010–2017)

Since 2011, OpenEdition, in partnership with the LSIS laboratory (UMR CNRS, Aix-Marseille University, University of Toulon), has developed a set of research and development projects around information sciences, automatic language processing and computer science to enrich the content of its platforms. The goal is to develop robust digital methods to efficiently search for content, suggest hypertext links generated by algorithms and provide reading pathways tailored to the needs of users.

Bilbo¹ , the cornerstone of the R&D projects

The first project focused on the automatic recognition and structuring of bibliographic references in scientific digital documentation, whether in the form of structured bibliographies at the end of the article, or footnotes, or quotations throughout the text. The identification of the references aims to enable navigation between different digital objects through the resolution of DOIs (Digital Object Identifiers)² .

Today, almost 12.7% of the bibliographic references on OpenEdition Journals, and 7.55% of references on OpenEdition Books, have a unique DOI.

Number of DOIs added on the OpenEdition Journals platform

Number of DOIs added on the OpenEdition Journals platform

The cornerstone of the OpenEdition Lab programmes, Bilbo, which emanates from this first project and is implemented on the OpenEdition Journals and OpenEdition Books platforms³ , continues to be improved through two other projects on content recommendation and the creation of links between OpenEdition documents.

Developing a reading recommendation system

In 2013, OpenEdition Center and the LSIS were awarded funding from the Provence-Alpes-Côte d’Azur (PACA) Region, the objective of which is the creation of a site dedicated to the automatic recommendation of books based on a dynamic analysis of book reviews in the human and social sciences. OpenEdition Review of Books⁴ is focused on linking books on OpenEdition Books and book reviews on OpenEdition Journals and Hypotheses, as well as on an extractor of noteworthy quotes – Echo⁵ .

The Inter-Texts Project⁶ is focused on the enhancement of documents through the development of software modules facilitating their interconnection, implicitly (for example, by their topic) or explicitly (by allusion, quotation, etc.). This involves designing and developing functionalities for aggregating scientific content, restructuring documents, validating sources, cross-linking content and a recommendation system. The research process also includes analyses on the layout of content, the textual content (semantic analysis) of its links (cross-references between documents) and uses (log analysis).

Lastly, two research projects run by the DILOH Equipment of Excellence, obtained in 2012, aim to produce a targeted recommendation system based on textual data and the automatic generation of hyperlinks.

Taken together, these projects have given rise to four theses led by PhD students from the LSIS (now the LIS, UMR 7020) and an internationally co-supervised thesis with the University of Vigo (Spain):

• Hussam Hamdan, Sentiment Analysis in Social Media, PhD thesis: Computer science. Marseille: Aix-Marseille University, 2015, 165 p. Viva: 1 December 2015.
• Chahinez Benkoussas, Approches non supervisées pour la recommandation de lectures et la mise en relation automatique de contenus au sein d’une bibliothèque numérique (Unsupervised approaches to reading recommendations and the automatic linkage of content within a digital library), PhD thesis: Computer science. Marseille: Aix-Marseille University, 2016. Viva: 14 December 2016.
• Anaïs Ollagnier, Combinaison d’approches du traitement automatique des langues et de la recherche d’information pour une recherche de livres orientée contenu efficace (Combining approaches to automatic language processing and information retrieval for effective content-oriented book searches), PhD thesis: Computer science. Marseille: Aix-Marseille University, 2017. Viva: 29 November 2017.
• Amal Htait, Modèles probabilistes de recommandation de livres par analyse de contenus et de sentiments, combinant Recherche d’Information et Traitement Automatique des Langues (Probabilistic models for book recommendation through content and sentiment analysis, combining information retrieval and automatic language processing). Adapted for the OpenEdition Human and Social Sciences series (in progress, viva scheduled for late 2018)
• Tamara Alvarez-Lopez, Sentiment Analysis in Social Media Content Using Technical NLP (in progress, viva scheduled for autumn 2018)

By focusing on some of the algorithms and utilities developed during these different theses, OpenEdition Lab has been able to classify and explore the contents of the platforms:

• extracting topics from the Calenda and Hypotheses corpora (https://github.com/OpenEdition/toolpic): highlighting topics or themes over a given period,
• categorization of particular types of documents such as book reviews (https://github.com/OpenEdition/toolpic).

We have also set up Language Detection to automatically detect the languages ​​of each document on the OpenEdition platforms⁷ .

Future OpenEdition Lab projects

Analysis of how readers use OpenEdition

Since 2017, OpenEdition Lab has extended its research activities to the ways in which users use the OpenEdition platforms. Funded by the PIA ISTEX programme, the Usages Alpha project consists in defining an analytical framework using a methodological mechanism, a set of usage models and the realization of a prototype of the unexpected reader detector that exploits the connection logs of the OpenEdition platforms. The study has involved large-scale quantitative analysis of complex data that is difficult to interpret. This research has focused on several elements:

• the modelling of the unexpected reader, as defined by Philip Bourne
• the classification of reading pathways
• a taxonomy of reading forms
• reading practices based on user profiles
• the observation of cycles of scientific output

Its results will be exploited as part of the Appropriations of Open Knowledge project funded by the Aix-Marseille University Foundation, A*Midex, from which OpenEdition has been awarded funding since October 2017⁸ . This project consists in understanding the uses of the OpenEdition platforms by seeking correlations between the content and the uses identified (for example, the long tail⁹ , the sleeping beauty¹⁰ , the unexpected reader¹¹ and the silent conversation¹² ). It will involve comparing the uses and the editorial and linguistic characteristics of the content to develop a typology of appropriations and the contexts of quotations in order to define scenarios for the propagation and appropriation of scientific information and to qualify the literacies of different populations of readers in the dissemination of scientific culture.

Rewriting the Bilbo code

A complete rewrite of Bilbo is currently under way. This aims to make Bilbo both more robust in production and more malleable as a research tool. It will enable Bilbo to be disseminated more widely to other corpora, to take into account larger volumes of data but also to be more easily integrated into a chain of processes such as language detection, adaptation of models, and searching for DOIs. At the same time it must be an easy-to-use development tool in what continues to be the very open field of extraction and annotation of bibliographic references.

Automatic indexing of books

In partnership with Huma-num, OpenEdition Lab is working on the indexing and categorization of all OpenEdition documents on the basis of different thesauri. A first step in this project is the automatic indexing of books.

The indexing of all documents on all the platforms will be a key intermediate result in the development of a reading recommendation system.

Reading recommendations for documents on OpenEdition

The More Like This project has just been launched on the Calenda platform. This project aims to produce relevant reading recommendations based solely on the Bag of Words in a text. Its integration into our database system has produced many successful tests. There is a possibility of extending these tests to all platforms, as well as integrating the project into a wider reading recommendation system.

Towards the opening of OpenEdition data

Finally, the OpenEdition Data project, which will start in 2018, will focus on opening the OpenEdition portal’s data. The latter represents an important resource for researchers and comprises a vast documentary collection, composed of a great diversity of editorial forms and languages. This project also signals OpenEdition’s desire to join the open data movement.

Footnotes

1. https://github.com/OpenEdition/bilbo [↩]
2. http://www.maisondesrevues.org/253 [↩]
3. http://lab.hypotheses.org/955 and http://lab.hypotheses.org/1625 [↩]
4. http://reviewofbooks.openeditionlab.org/presentation [↩]
5. https://github.com/OpenEdition/echo [↩]
6. http://grapher.openeditionlab.org [↩]
7. Sheeren Albitar, “À la recherche d’un détecteur automatique des langues”, Carnet d’OpenEdition Lab (research blog), 16 November 2015. URL: https://lab.hypotheses.org/1048 [↩]
8. http://lab.hypotheses.org/1845 [↩]
9. Anderson, Chris. 2012. The Long Tail: Why the Future of Business is Selling Less of More. New York: Hyperion. [↩]
10. Van Raan, Anthony F. J. 2004. “Sleeping Beauties in science”, Scientometrics. 59: 467. https://doi.org/10.1023/B:SCIE.0000018543.82441.f1 [↩]
11. Kevin Smith, J.D.. 2011. “The unexpected reader”, Scholarly Communications@Duke. https://blogs.library.duke.edu/scholcomm/2011/11/15/the-unexpected-reader/ [↩]
12. https://archivesic.ccsd.cnrs.fr/sic_00439849 [↩]