Asterisk1d ago

I have a problem talking about #openaccess when it comes to the usage of #transkribus.

I have also a problem when using #AI generated images in research presentations.

Please don't use Transkribus when you want to commit to #OpenScience but rather #eScriptorium or similar.

Please don't use #AI generated image slop.

Thank you. #dh2025

Show threadFrederik Elwert1d ago
@jomla Is it now consensus that Transkribus is one of the bad guys? I was always torn – I am critical of the closed-source nature of some parts, but appreciate the cooperative structure. I had hoped the new Open Source engine (pylaia, I think?) would improve things? But I haven’t checked in for a while, so I might have missed the recent developments. (You’re not the only one who I heard take a critical stance lately, so I’m genuinely interested.)
Show thread@frueheneuzeit1d ago
@felwert @jomla well, we also all use Oxygen. This is strictly my personal opinion, so please don't tar and feather me: technology is about trade-offs. As is openness. And if the additional efforts of not choosing the easy solution lead to an outcome where people don't do the work they wanted to do, I'd probably draw a line.
Show threadAsterisk1d ago
@stefan_hessbrueggen @felwert I don't really have an issue with people using #Transcribus, it is a powerful and easy to use tool for all kind of challenges. As @christof sais, it is scholar-led, it is a cooperation, it is not "bad guys", I'd say. But I indeed have an issue with people depicting it as "open", as it rather follows a policy against that. No way to share models, kinda-locked-in architecture, credit-based access, all that is not "evil", but definitely not #OpenAccess.
Show threadV ¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯1d ago
@jomla @stefan_hessbrueggen @felwert @christof Looking at Transkribus's own terms: they 'retain all rights, including Intellectual Property Rights' over core algorithms, can 'modify Products and Services at any time and terminate operation,' and everything runs through their Oracle Database backend. So while they market accessibility, the foundation relies on expensive proprietary database tech with zero user control? For open access advocates, this may read as a form of gatekeeping, doesn't it?
Show threadAndreas Wagner19h ago

@vmmmh @jomla @stefan_hessbrueggen @felwert @christof I am not sure these points are so critical. What are we assessing the openness of, after all? The core ATR software, the models, the scholarly initiative, the platform instance at that URL, the software to run the platform *around* the ATR software?

- claiming IP over core algorithms is no hindrance to publishing them under an open license
- reserving the right to modify products and services and terminate operation might be reasonable? Depends on the contracts they have with their subscription customers I suppose. At least I think it is very common for platform service providers.
- using proprietory software to organize the platform backend db: As long as this does not affect the license of the core ATR processing components I really don't think this is a problem.

It's a different thing with not publishing the core algorithms, models and workflow openly and transparently (FAIRly), however. Here I agree it could be much better.

Show threadV ¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯
@anwagnerdreas @jomla @stefan_hessbrueggen @felwert @christof Openness can evolve with user needs, sure—and maybe it’s less about any single component, and more about whether diverse communities can shape, adapt, or migrate their work if their context or priorities change. I wonder if clear “exit strategies” and open data standards are as important as code openness for building resilient, future-proof research environments. Transparency in these areas could help everyone win.
Jul 18 at 8:42amElk