Grants talk:Programs/Wikimedia Community Fund/Rapid Fund/WikiTermBase 2.0 (ID: 23553206)

Recommendation to fund

Hello Michel Bakni, thanks for submitting this Rapid Fund proposal. We finalized the proposal review process and we are sharing with you the summarized recommendations from the review team.

We believe this proposal is well planned and responds to a clear community need. It builds on an existing Toolforge service and an already functional gadget. There seems to be active interest among Arabic Wikipedia users, and it is positive to see a small team of three developers supporting it.

We note that two proposals from this Rapid Fund cycle: WikiTermBase 2.0 (R-RF-2508-20079) and TermBox project (R-RF-2509-20123) are effectively similar proposals aimed at different wiki language versions. While each proposal is interesting on its own, we request that both teams collaborate so the impact could be greater with coordinated efforts. We also request that the team aligns its strategy with the Content Translation team efforts to avoid any duplication or conflict. From a product standpoint, it is important to ensure this tool does not conflict with or duplicate functionality of the Content Translation tool.

We recommend this proposal for funding with a few follow-up questions:

• It is unclear what challenges exist to make this a default gadget. What conditions the tool needs to meet in order to hit such a threshold?
• How extensible is WikiTermBase to other languages beyond Arabic? How will you make sure that it can be also useful for other languages?
• The stated goal of achieving a 20% reduction in translation time would benefit from clearer definition. How will the 20% time-saving goal be measured? And does this represent a total 20% reduction from the baseline, or an additional 5% improvement over the previously reported 15% time saving?

On behalf of the review team, DSaroyan (WMF) (talk) 07:50, 8 October 2025 (UTC)

Well-received, replay in several days. Michel Bakni (talk) 11:08, 9 October 2025 (UTC)

1. Re: Default Gadget requirements: The criteria for a default gadget are laid out here (the local Arabic Wikipedia criteria are very similar). Frankly, we ourselves are still figuring out how to navigate them exactly, as they are a bit vague and qualitative. What we understand so far is that making a gadget default isn't easy, and will at least require proving that we have very robust tech. That's why we're planning on investing significant effort into both widely testing our tool, as well as engaging stakeholders who would be involved in such a decision, with a focus on the Arabic community (who have the mandate to enable the tool locally), and the Wikimedia Foundation's Language Team (who best know related technical infrastructure on the Foundation side).
2. re: Supporting other languages: Our infrastructure is language agnostic. The tool's code is fully available on Github, and can be used directly to save a team like the Vietnamese Termbox team the majority of the front end and back end development. However, we do note that a significant portion of our work has gone into the content hosted in our database, which we meticulously collected from public domain dictionaries (an approach similar to what's been proposed by the Termbox team), and this content is mostly language specific.
3. re: Measuring the translation time goal: The goal is to reach 20% total time saving consistently reported from the tool. Indeed, our past data has indicated around 15% reduction, although admittedly from a small sample. We aim to gather more samples and keep targeting 20%. How this has been measured in the past is by self-reported estimations from a user survey. While we will continue to keep gathering survey input (since it’s the easiest format), we do hope to conduct thorough live user experiences to confirm these numbers, should we hopefully find people willing to voluntarily participate.
4. re: Engaging with the Content Translation Team: We have been engaging the Foundation’s Language Team since our ideation phase, almost two years ago now. We did discuss the potential for integrating such a tool into Content Translation in detail with staff including Niklas and Amir, on multiple occasions. Although the conclusion was that an integration would be very challenging, we did get several useful ideas from them (e.g. connecting the tool to Wikidata Lexicon). We definitely plan on continuing to keep them in the loop.
5. re: Engaging with TermBox: We directly reached out to the TermBox team and have offered a meeting to share our experience and learn more about their work. Their work is definitely very similar, and our lead programmer is willing to share code and also learn more about their plans.

Michel Bakni (talk) 10:45, 12 October 2025 (UTC)