Autre Publication Scientifique Année : 2025

Referee report. For: The artificial intelligence cooperative: READ-COOP, Transkribus, and the benefits of shared community infrastructure for automated text recognition [version 1]

Résumé

The paper presents, explains and analyzes the choice and its aftermath behind the structural change from a publicly funded project (Transkribus) to a cooperative-based ownership of the infrastructure and developments. It uses the authors' own experience and their own company as a use case to defend the hypothesis that cooperative structuration in niche AI applications—but even more for research projects after public funding phases is a way to ensure its survival and sustainability. The authors present the cooperative status of READ-COOP as the element responsible for the success of the platform (Page 2, "This success [number of documents and users] is due to the cooperative approach..."). After carefully reviewing the paper, I find that it effectively presents the argument that the cooperative structure has, for the time being (as the authors rightly acknowledge on page 29), ensured the short- to mid-term sustainability of the tool originally funded by public funds. However, the paper lacks sufficient argumentation and supporting sources to substantiate the claim that the company structure is responsible for the platform's success (measured in terms of the number of users, documents, or citations, as referenced in section 5.6 by the authors themselves). Furthermore, the paper appears to omit critical perspectives on this structural change and fails to appropriately cite the work of competitors. At times, the paper reads more as an advert for the cooperative (for example, the fact that Transkribus won an award is mentioned three times and cited twice in the bibliography) rather than a comprehensive analysis of this structural shift. In the following report, I will suggest additional avenues that I believe the authors should consider integrating into their paper in order to better represent the impact of the cooperative structure, its benefits, and its consequences on the community. Finally, the very small participation rate in the survey, along with its imbalance toward “founding” members (35% from 2019 and 56% from 2019 and 2020), warrants a more cautious interpretation of the results. For instance, when analyzing responses indicating that the primary motivation behind shareholding was the discount, it is important to consider how many of these respondents are recent members. Conducting an additional survey focused specifically on users would provide valuable insights and further strengthen the overall claims and conclusions of the paper. If institutions and individuals become members for discount rather than for cooperation, the same way they pay for subscriptions, how does the argument of the paper hold? I would like to conclude the introduction to my review by expressing my appreciation for the transparency demonstrated in this paper and its intent to provide a detailed perspective on the transition process that Transkribus underwent. It is undoubtedly challenging to write about the success of one’s own project, especially when it has been nurtured for over a decade. However, this close connection to the project can—as is the case here—limit the necessary critical perspective on practices that are genuinely aimed at supporting the community but also function as significant marketing and user retention strategies. This paper should be revised and resubmitted to address the issues I think are too big for a "Approve with reservations".

Fichier non déposé

Dates et versions

hal-04957346 , version 1 (19-02-2025)

Identifiants

Citer

Thibault Clérice. Referee report. For: The artificial intelligence cooperative: READ-COOP, Transkribus, and the benefits of shared community infrastructure for automated text recognition [version 1]. 2025, ⟨10.21956/openreseurope.20281.r50056⟩. ⟨hal-04957346⟩

Collections

212 Consultations
0 Téléchargements

Altmetric

Partager

  • More