[Wikimedia-l] Re: Wikifunctions needs more love!

Hi Noah, Thank you for your interest in Wikifunctions! And I agree, I would love to see much more love for WIkifunctions :) You’re asking exactly the right questions, and thank you for raising them here, as it allows us to refine the FAQs and the documentation for Wikifunctions, so that it becomes hopefully clearer for people in the future who make the same journey of discovery that you had. Answering your questions requires me to make a few subtle differentiations, and I apologize for the lengthy answer. 1) Why is Wikifunctions not an “encyclopedia of algorithms”? The Wikimedia movement already has an encyclopedia of algorithms. It's called Wikipedia. Wikifunctions is a platform with which the Wikimedia community will, in time, be able to use functions to build encyclopedias, text books, interactive content, and much more. We're not interested in functions for their own sake, or just as a way of documenting interesting code. We're looking for functions that are valuable for the Wikimedia's community's content. Some educational content could indeed be a comparative gallery of different algorithmic approaches to the same broad problem, but that's a limited area of interest compared to natural language generation, data processing, and the general computational needs that our communities have already shown the need for. Indeed, Wikifunctions makes a distinction between a function – which answers a question – and its implementation – the mechanics of creating an answer for a question. Algorithms are directly relevant for an implementation. The focus of Wikifunctions is on functions, not on implementations and thus algorithms (although the latter are obviously crucially needed for the former). We are worried that by calling ourselves an “encyclopedia of algorithms” we would create expectations that one should, or can easily, capture a diverse set of algorithms in Wikifunctions. But there will be many algorithms that we won’t be able to represent given the intentional system design trade-offs (e.g. Shor’s algorithm - https://en.wikipedia.org/wiki/Shor%27s_algorithm ), and a lot that are not immediately useful in Wikifunctions. This is why we prefer to present Wikifunctions as a library of functions instead. The difference is subtle, but we hope it makes sense. We are not against people being motivated by providing a set of standard algorithms through Wikifunctions, but be warned that they might find themselves being somewhat frustrated because the requirements for such a goal will not be prioritized by the development team. We are happy to discuss this point more, as it helps to sharpen our shared understanding and wording around the goal of Wikifunctions. 2) Why are pretty much all the functions trivial? I am sure it is because we just started. Public availability of Wikifunctions only opened two months ago, and we have more than 600 functions available. It is to be expected that many of these early functions will be also available in standard libraries, particularly of languages such as Python which are known for having expansive standard libraries - https://xkcd.com/353/ Having said that, you point to the “70% of natural language generation functions”. These are in fact often not covered by standard libraries, and not just wrappers around existing standard library functions. You can provide more than one implementation in one language, i.e. there being an implementation of SHA-256 using the standard Python hashlib does not mean you cannot implement it in Python without using the function from the library. That’s totally OK. Just no one did it yet. 3) Why are all the functions named something like `Z10124(Z10124K1)`? They’re not. That’s something we need to improve in the UI. The name of the function with the ZID Z10124 - https://www.wikifunctions.org/view/en/Z10124 - is, in English, SHA-256. Each function has names in different languages. We still have a few spots where the IDs are directly visible. We want to reduce the number of such spots. If you would like to help us with that, you would be very welcome. 4) Why are there so many natural language functions? Because one of the main goals of Wikifunctions is to provide the functions necessary for Abstract Wikipedia. 5) What is meant by this? Abstract Wikipedia, as well as why it is built on top of Wikifunctions, is described further here: https://cacm.acm.org/magazines/2021/4/251343-building-a-multilingual-wikipe… In short: we are aiming to use natural language generation functions in order to create accurate Wikipedia articles in many different languages from a single abstract representation about a topic. These articles can be used to fill current gaps in many language editions of Wikipedia. At the same time, the abstract representation would be maintained only once, but automatically kept up to date in all participating languages. x) Why do we only have functions for strings, booleans, and lists? Because we are still working on supporting more types. You can follow our progress through our weekly updates: https://www.wikifunctions.org/wiki/Wikifunctions:Status_updates https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Updates We are working right now on the features necessary to support our next type, numbers. That should allow for many more algorithms and functions that will need to be created and maintained. 6) What can I do to help it reach its full potential? Join us! Join us in working on the functions we have, suggesting and creating new functions, creating more tests and implementations. Join our community in our discussions. Join us as a volunteer contributor to the codebase of Wikifunctions itself. If you need pointers on how to join us, let us know. I hope this clarifies a few things! Cheers, Denny On Sat, Feb 10, 2024 at 12:31 AM noah poulton <noahpoulton2(a)gmail.com> wrote: