[DOCS] Create information architecture draft for docs

[jessicarose]

Topic of request
What topic does this documentation request concern?


Currently documentation is spread across several sources and difficult to maintain, find and change.

I think researching and creating an information architecture plan is a good next step to approach documentation in a strategic fashion that works best for the different types of dataset users, contributors and other stakeholders interested in Common Voice.

Kind of request
Is this a request for new documentation? An update? Or a change to existing docs?


Other: this ticket is to create an outline of the current documentation and new documentation needed, segmented by use case or user/contributor persona.

If this is a request to create new documentation, what do you think is important to include?

If this is a request to update existing documention, what needs to be updated?

If this is a request to change existing documentation, what needs to be changed?

[HarikalarKutusu]

For what it's worth, in the past while working with Hillary, we talked about this a bit. In short, here is what I proposed:

• Make use of readthedocs (it is free for open source, support github integration, document versions and multi-lingual documentation):
  • Use a Github repo common_voice/cv-documentation and link it to readthedocs
  • Create multiple localized branches, make locale leads moderator, but keep final control for formatting etc. The content (translation) should be under locale moderators.
  • You update the documentation through Github workflows (issues & PRs)
  • Whenever you need to mention something in WebApp, Discourse, Matrix, just point to the documentation. That would help support the load a lot.
• Divide the documentation as follows:
  • General info on what is MCV, release timing, etc
  • Info on AI, Voice AI, how it is done (from text corpora to models - short info and external links)
  • User documentation (for WebApp - which is where)
  • Community Leads documentation (mostly from the community handbook)
  • Dataset user documentation (structure)
  • Developer documentation (how it works internally)
  • Terminology (needed to explain concepts like "variant" to the general population)
  • FAQ (for questions like "How much would I record", and "How can I add bulk sentences" - add "see: link" for these, ...)
  • Future development & how to contribute to code
  • etc

I kept a rather detailed documentation in Turkish in discourse and try to keep it up-to-date:
https://discourse.mozilla.org/t/surec-dogrular-yanlislar-ve-veri-kumesinin-iyilestirilmesi/85938
You may like to g-translate that to see what I mentioned. Except for causal users who record 5-10 sentences, you need to give them introductory info on AI and voice-AI to show them why it is the way it is. You'll also see locale-specific information, such as where to check for spelling, borrowed words in that language, etc - that is the reason Pontoon fails and you need multi-lingual branches on github.

I hope this helps...

Edit: Added "Terminology".