Community Wishlist Survey 2019/Editing/Make Wikipedia more accessible to the visually impaired

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search
Random proposal ►

Support ◄ Back to Editing


  • Problem: So far, editing/using Wikipedia or other Wikimedia Foundation projects require full usage of our visual, because of keyboard-input method. We need to find a way to expand Wikimedia Foundation project interface to be able to accommodate the visually impaired people so that they also can get the benefit of this free knowledge for mankind of Wikipedia. (I rewrite this statement from the previously written at Community Wishlist Survey 2019/Programs and events section)
  • Who would benefit: People having difficulties in their visual in using Wikipedia.
  • Proposed solution: Make Wikipedia (and other Wikimedia Foundation projects) to be visually impaired-friendly. First, for searching the articles in Wikipedia, we need to create a sound-sensitive input method so that they can easily search any article by speaking (e.g. into microphone input) and automatically convert it into wording and automatically search the related article names from it. Second, for each article, there must be a button that can be clicked (or sound activated mode) in which it will then read out all of the wordings inside one particular article to the user.
  • More comments: So far the closest works for this kind of project is Wiki in Audio (including Wikimedia:Spoken articles, Wikipedia:WikiProject Spoken Wikipedia, Meta:Wikisound). But what I have understood so far, those sound must be fully recorded, thus disabling the spoken voice to speak out when the article gets expanded. So, it is better to make each of the word to be spoken one by one, instead of recording a voice for the whole article continuously. Maybe a more similar approach would be like the voice button at Google Translate, in which it will speak out each word one by one.
  • Phabricator tickets:
  • Proposer: Chongkian (talk) 04:12, 6 November 2018 (UTC)

Discussion[edit]

  • The reason I don't think this is possible is that speech-to-text and text-to-speech software is subject to patents held by various companies. The software that runs Wikipedia is required to be open licensed just like the content. Patents last about 20 years, so speech-to-text and text-to-speech technology that is not restricted by patents is from 1998 and so not very good. Also, just to let you know, w:Wikipedia:Manual of Style/Accessibility is a guideline in the Manual of Style that covers accessibility. AHeneen (talk) 08:32, 6 November 2018 (UTC)
    • @AHeneen: I partially disagree. Free text-to-speech software and technology has existed for ages (Orca comes to my mind as an example). I don't know about speech-to-text though. --AKlapper (WMF) (talk) 11:37, 6 November 2018 (UTC)
      • AKlapper (WMF), Gnangarra: For copyleft and other open-source voice-to-speech software, see Speech recognition software for Linux. If you can speak or understand speech, Voxforge is copyleft and could use your contributions. The voice-assistant en:Mycroft (software) was a copyleft version of Cortana/Alexa etc., but then they relicensed it, and I don't know of a copyleft equivalent. HLHJ (talk) 01:44, 18 November 2018 (UTC)
        • @HLHJ: there's a need for it to be incorporated somehow into the mobile Wikipedia app, with step like pending changes for the obvious spelling/grammar reasons. Gnangarra (talk) 03:28, 18 November 2018 (UTC)
  • For the second part of this proposal (text-to-speech), see mw:Wikispeech. Also see phab:T194014 about making Wikispeech available in Wikimedia projects. --AKlapper (WMF) (talk) 11:37, 6 November 2018 (UTC)
  • Not really needed; blind people like me use screen readers, and helping those who don't is way way beyond the scope of what almost 99.99999% of websites do. Graham87 (talk) 11:50, 6 November 2018 (UTC)
  • @Comp1089: may be interested. --Tohaomg (talk) 15:52, 8 November 2018 (UTC)
  • I also use a screen reader, so the text-to-speech thing is not an issue in most cases. However, text-to-speech (input and/or reading) is not supported in a great number of languages, which is a real problem. For instance, due to the unavailability of a voice able to read Bashkir (or Kazakh) Cyrillic script, I always have to write the text in Latin transliteration (and then convert it to Cyrillic, if it is a Wikipedia article). Advocacy for creating/developing screen reading voices for as many languages as possible will probably be a task of the Para-Wikimedians Community User Group.
    As far as speech-to-text is concerned, enabling all people to search through the content using their voice seems a rather good idea. This may also come in handy when a blind/visually impaired user is for some reason unable to use a keyboard. (On a side note, there also are sighted people who cannot do that either, e.g. due to movement impairment). However, introducing such a technology to all Wikimedia wikis will definitely require a lot of time and other resources. --Comp1089 (talk) 19:05, 8 November 2018 (UTC)
  • Voice search does already exist in the Android and iOS Wikipedia apps via standard integration with their operating systems. However, very good points above about limited language support, and those apps only help on the Wikipedias. Quiddity (WMF) (talk) 22:16, 8 November 2018 (UTC)
  • observation isnt the fact that all speech to text and text to speech are propriety software as reason to develop a open source version so that everyone can share in the sum of all knowledge, I assume the visually impaired fall into the subset of everyone, also as tool it could help with contributions from smart and mobile devices. Gnangarra (talk) 01:17, 9 November 2018 (UTC)
  • @Graham87: Sorry for abusing this thread for a more specific question. I would be interested if you can understand Wikipedia pages with mathematical content. I guess mathematical articles that make heavy use of template-hacks are the worst, I would guess most of them are not really accessible at all. However also for math formulas using the normal <math> tags, speakText generation is disabled (phab:T120938), so the screen-reader would have to pick up the hidden MathML and generate text itself. Does this work well? Is this sufficient to understand more complex formulas, or do you need the interactive "accessibility explorer" that gets shipped with normal MathJax and pages like math.stackexchange.com use? At least for someone like me who is not used to screen-readers, this accessibility explorer where you can navigate with arrow keys is the only way how I might understand a more complex formula.--Debenben (talk) 19:49, 12 November 2018 (UTC)
I think I may respond here either. I do experience difficulties reading the more complex formulæ. The solution you suggested seems to work best with NVDA. @Graham87:, which screen reader do you normally use? --Comp1089 (talk) 21:02, 12 November 2018 (UTC)
I use JAWS with its inbuilt math viewer, which works well for simple formulae (and is slow but eventually works for more complex ones) ... it's usually good enough for me. Graham87 (talk) 01:28, 13 November 2018 (UTC)
Hi, @Chongkian: As was mentioned in the discussion, building a voice-to-text (and a text-to-voice) system is quite out of scope for the Community Tech team, and there are already several tools that already exist that are providing very good solutions. The problem statement itself -- improving the accessibility functions of Wikipedia for the visually impaired -- is extremely important, and we will be happy to tackle that; if the wish is voted into the top 10, we will be doing an accessibility audit on Wikipedia's main functionalities and tackle the necessary fixes, while scoping the size of the work to be realistic for the team to achieve as part of the survey. Thank you for participating in the survey! MSchottlender-WMF (talk) 01:31, 14 November 2018 (UTC)
Sounds good. Is there any way to make those graphs for which we have the raw data accessible to the blind? Graph Template Collection] graphs, for instance. Separately, how about having colour filters for the colourblind, who often can't read graphs and diagrams (mentioned here)? Or at least a way to auto-tag problematic images... HLHJ (talk) 08:02, 14 November 2018 (UTC)
I believe we all should do this one by one, and step by step. Of course we can't have everything that is accessible to normal people to become accessible to the visually impaired. We can always start first with the text to audio reading thing, including all of the article searching, and probably searching of categories or even switching between different languages (of course we shall start from the major world's languages first). And then we shall see how we can "visualize" (in a broader sense) on all of those visuals (picture, graphs, videos) to the visually-impaired people. Those with full or partial color blind, probably we should have some equivalent (either upload it double-ly/redundantly (how to put this into words? - like the ability to easily switch between visual editor or normal coding) black & white graphs/images, but with color description (pointers telling this section is what color & that section is what color). Or maybe in the future, every image uploaded to Commons, we can always give the voice input (beside naming typing input), so that the uploader or other Wikipedia users can record the voice (or type the description and convert it into voices) to describe the photo (e.g. "A photo with a kid on the left, sitting looking into a pond with 5 ducks swimming towards him during late afternoon."). Basically it goes back to our childhood when our mothers read a storybook to us before we go to sleep, or during the era of radio before television appear, where people fully depend on sound for "visualization" when hearing a story (equivalent to "watching movie" nowadays) broadcasted from radios. Chongkian (talk) 07:32, 15 November 2018 (UTC)

Hi all, thank you so much for all of the great & beautiful response :') Chongkian (talk) 07:27, 15 November 2018 (UTC)

Also check out this proposal..... *excerpt*

Problem: Available Text to Speech for all wiki Articles (Wikispeech extension). Let the articles talk when Required.
Wikispeech logo.svg
Proposed solution:
1. "Wikispeech" is a text-to-speech tool (in Beta) to make Wikimedia's projects more accessible for people that, for different reasons, have difficulties reading. At first make it available for en wikipedia . then asign more developer so that it can have cortana/siri/google assistant like "precise accent" & "natural voice".then make this extension/api integrate with mediawiki software. And please make the voice more natural like this "Cortana,Alexa" or "Google Assistant".
2. Google text to speech api is licenced under apachi 2.0 . Is it possible to ask them to make a github fork with CC/GFDL licence only for official wiki project, cause this proposal also humanitarian? They have natural-sounding speech with 30 voices, available in multiple languages and variants.

—The preceding unsigned comment was added by Ahm masum (talk) 20:07, 15 November 2018 (UTC)

Voting[edit]