Let’s Publish a Knowledge Base from Ask Fedora

Photo by Kaja Sariwating on Unsplash cropped

“We often get questions..”

Just like bug reports or support tickets, a user Q&A support forum is a gold mine for understanding users and contributors.

Newcomers get help from contributors to solve problems in user Q&A support forums like Ask Fedora. Some of the solutions work for many situations where users can apply the solutions without modifications. To our surprise, there are repeated questions for similar or related issues every week.

Content reuse and rewriting

This is where content reuse comes in. The idea is to select top posts in Ask Fedora and consolidate them for How-To and FAQ-style user documentation. Why? Content reuse and rewriting can help make maintenance and enhancement of articles easier than in forum posts. With this in mind, I gave a talk about “Documenting Top Answers in a User Support Forum” at DevConfCZ in June 2024.

Call for Participation: Reviewers and Writers

We are looking for anyone interested in this effort to join us.

To kick off this effort I analyzed forum posts by pulling a list of tags that relate to multimedia. Next I grouped them by top answers on multimedia.

How to debug sound problems
Audio input not recognized
Use of multiple Bluetooth device
PulseAudio and PipeWire troubleshooting

Docs team have contributor guides for submitting your draft to a Quick Docs repository. If you need help with the process, please stop by at the Matrix Docs room for more informal communication. Thank you!

Fedora Contributor Community Fedora Project community

11 Comments

  1. F Hansen

    How about first just try to train an AI on the dataset?

    • Darvond

      You would trust an internet connected parrot to reasonably produce results that are legible, accurate, and up to date?

      AI is horse hockey. You can’t just throw it at complex tasks like this.

    • I got help from the Discourse AI plugin to to summarize topics – often walls of text in Ask Fedora 🙂 that work alright. However, it doesn’t not accomplish final tasks like;
      1) assess accepted solutions (‘Solution’ box ticked): there are different approaches to a problem, but AI throws out them in a sequential order.
      2) shortlist basic steps vs advanced troubleshooting: there is no differentiation in initial AI output about when we need logs or not.
      3) summarize related posts with the same tags
      4) review the results before publication
      5) encourage more people to maintain articles and enjoy writing

      We can’t delegate these human factors to Open AI yet.

      The main driver of content conversion/reuse is to drive down repeated questions in Ask Fedora and improve user experience of multimedia users at Fedora Linux as a start.

      I wanted to start with a topic, multimedia, first and then encourage more subject matter specialists to join and write about different topics. It is more about ‘documentation culture’ with this initiative, not delving into what AI could do.

      • Zoltan

        Question: Would be possible to integrate an AI with a chatbox in the future? IT would be much natural to talk directly to an AI like chatgpt.

        • What problems have you identified from knowledge creation process in the Fedora Project? I can’t see your question is really relevant to this article.

          • I suspect Zoltan is not referring to the creation of the knowledge, rather the search for and access to the correct problem resolution once the knowledge has been gathered.

            • highhills

              Knowledge creation workflow in AI is actually in sync with summarizing results of ChatGPT and LLMs, which don’t deliver accurate context-aware information. Troubleshooting specific issues yields unreliable results.

              Information in Ask Fedora might not help curating good results with ChatGPT and LLMs because if there are conflicting solutions for the same issue, the AI would spit out not-so-accurate summaries.

              Many of informed users of AI in profession knows of hallucinations with ChatGPT and LLMs and are undeniably pessimistic at times. When people recite AI without context in real-life, they are faced with pessimism.

              Some don’t ask right questions.

            • Zoltan

              You are right. The problem is that almost anything can be a knowledge, and data that can be important to create a result. Kinda chicken or egg question. Because AI will transform everything, not just search. Linux has the strongest base, but somehow we forgot to use it. What I mean here is the granular modularity between apps, and build up from there. More widely, we don’t use the one program – one purpose anymore, instead we use monolithic approach to tasks that we wish to solve our problems. Back in the old days linux had a chains of commands with pipe and more – it was fast, and effective. Now, with AI we can do that again, just faster, with the right modules to get the best results. Imagine an AI as part of your system, that uses ML after your needs, and requests. The knowledge is that how you as user want to do, and what to do. That means we have to gather some telemetry to feed the ML, and the AI, then we can offer apps/solutions that only uses a single universal skeleton and bunch modules basically from any source what is modify/adds functions to it. That will give you huge scaling ability, and with grouping (aka kits), and combination with AI, that know specific ways to render on the fly an app, render a macro button to your screen, or simply answer your question – the key will be not the data, else the modular integration of the built by AI system. Data can be anything, the source can be anything both physically and logically too, but the path by Machine learning (ML), and proper extrapolation method with AI will change computing as we know in long term.

        • Dave Grossheim

          Ask Fedora has a Topics (AI Search) option in Semantic Search, but not integrated with like a chatbox. Caveat is Discourse AI Search does not always find topics with 100% relevancy. It is experimental.

          You could ask the question on this thread.

          https://discussion.fedoraproject.org/t/more-ai-is-here-new-open-source-powered-features-in-testing-on-fedora-discussion/82655

        • I;m Sorry I misunderstood your question. I have an answer for your question now.

          Check this link below.

          https://discussion.fedoraproject.org/t/discourse-ai-plugin/126191

          In short, we haven’t enabled AI Bot after much deliberation. I’m in favor of best use of self-help with AI. We’ll be exploring options when the solutions fit the purpose in the future.

  2. Piotr

    Zoltan, there are freely available AI tools in the form of chatbots that allow you to search for information on specific websites or a list of websites based on the links you provide. You can simply give Gemini or Perplexity a link to the Fedora forum or Fedora documentation site and ask any question, explicitly requesting that the answer be based on that specific source.

Leave a Reply


The interval between posting a comment and its appearance will be irregular so please DO NOT resend the same post repeatedly. All comments are moderated but this site is not monitored continuously so comments will not appear as soon as posted.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

The opinions expressed on this website are those of each author, not of the author's employer or of Red Hat. Fedora Magazine aspires to publish all content under a Creative Commons license but may not be able to do so in all cases. You are responsible for ensuring that you have the necessary permission to reuse any work on this site. The Fedora logo is a trademark of Red Hat, Inc. Terms and Conditions