Jump to content

Community Wishlist Survey 2022/Larger suggestions/Spoken Articles

From Meta, a Wikimedia project coordination wiki

Spoken Articles

  • Problem: Projects to make Wikimedia articles spoken have not gone ahead and/or the number of audios compared to the number of articles is negligible. This is caused by two reasons: Few editors know how to make these audios and over time these audios become outdated by the edits made.
  • Proposed solution: One of the possible ways to expand and improve this idea is to make software that reads (with pre-established rules) the text of articles in all Wikimedia projects in all languages automatically, allowing people to choose to listen or read, observe misspellings more easily to correct them, learn the pronunciation of words and the visually impaired can benefit from this.
  • Who would benefit:
  • More comments: You might find interesting pronunciation lexicons for XHTML and SSML attributes in XHTML. There could be wiki syntax, possibly templates, for listing pronunciations which could be aggregated into pronunciation lexicon resources referenced with document metadata. There could also be one or more wiki templates for rendering XHTML spans with SSML attributes. As for which document content to synthesize, this could also be done using cascading stylesheets using the CSS Speech Module, specifically the speak property. (AdamSobieski idea)
  • Phabricator tickets:
  • Proposer: N4CH77 (talk) 18:56, 10 January 2022 (UTC)[reply]

Discussion

Agree I think this would be a great idea. This would surely help out those with visual impairments. Although they have TTS, these can often mess up the formatting and read unnecessary information (such as citations) in a way that would confuse a listener. Perhaps before software is made, users can upload audio of them just reading articles? Jmaxx37, 02:10, 11 January 2022 (UTC)[reply]

Spoken Wikipedia exists for just this purpose currently. Agree that automation would be great if possible. Retswerb (talk) 05:11, 11 January 2022 (UTC)[reply]

Are you aware of Wikimedia Sverige's Wikispeech project? Jon Harald Søby (talk) 10:01, 11 January 2022 (UTC)[reply]

Yes, I'm aware. But this project only has 4 languages available and even with a page explaining it, I didn't understand how to download it. (sorry)
Maybe if I get more stuff it can be used to read the text automatically. --N4CH77 (talk) 12:11, 11 January 2022 (UTC)[reply]
  • Browsers have an entire Speech synthesis api these days.. Why are we not using that ? Seems simpler/more durable. Also, I can just select a page in my browser and have the browser read it to me. Been an option for about 8 years now I think. On my phone i can select text and choose "speak". Rebuilding things that the OS or browser can already do is generally a bad idea, we are NOT better at it than Apple or Google. —TheDJ (talkcontribs) 12:44, 11 January 2022 (UTC)[reply]
At least, for me, these browsers don't work the right way (Reading unnecessary parts, wrong pronunciation, reading in a disordered way and in quotes and in images they don't say what it is, it only reads the text below thus harming readers and the visually impaired). I don't know what a new software idea that reads with pre-established rules and correctly has to "bad" and at no time did I say that you are better than Apple or Google (and why this is taken into question ¯\_( ツ)_/¯) --N4CH77 (talk) 15:55, 11 January 2022 (UTC)[reply]
This could probably be written reasonably well for languages like Spanish, where any spelling has precisely one way it could be pronounced. Writing it for languages like English, where there are a lot of exceptions, would be much more difficult, and writing it for languages like Hebrew, where vowels are frequently omitted, is probably impossible. Animal lover 666 (talk) 19:47, 11 January 2022 (UTC)[reply]
It doesn't have to be done all at once (especially since I know it will take time to program these languages) and the pronunciation doesn't have to be THE MOST PERFECT in the world, it just needs to be understandable —N4CH77 (talk) 20:38, 11 January 2022 (UTC)[reply]
I still think that this would be impossible for languages like Hebrew and Arabic (although simple for Yidish, which has a 1:1 correspondence for words which aren't from Hebrew or Aramaic) and very difficult for languages like English and French. Animal lover 666 (talk) 09:23, 12 January 2022 (UTC)[reply]
Text to speech engines do the work. Most of the time they work fine. Just copy any paragraph into Google Translate and click the Play button to listen to how they pronounce it. It is indeed troublesome for languages like Japanese where a single words have multiple possible and distinct pronunciation depending on context, but most modern text to speech engine have reasonable accuracy rate at guessing which pronunciation to use, according to my understanding. And they would work much better than anything developing from new by the community tech team. C933103 (talk) 00:24, 16 January 2022 (UTC)[reply]
The reading incorrect parts thing and the wrong order thing can probably be improved by improving accessibility of Wikipedia's website, allowing the website to properly tell the browser which parts should they read. And I think it would also be much easier to done and have much higher chance to accomplish than creating an entirely new software just to read Wikipedia. C933103 (talk) 00:30, 16 January 2022 (UTC)[reply]
I agree, it's much simpler to do and would improve screen reading without much effort. —N4CH77 (talk) 19:21, 16 January 2022 (UTC)[reply]
  • This should probably be in the Multimedia and Commons category. --Izno (talk) 22:48, 18 January 2022 (UTC)[reply]
  • Hello and thanks for taking the time to write this proposal. The needed support of voice media to include and disseminate more knowledge makes complete sense. However, we reviewed this proposal as a team and have determined that this is out of scope for our team due to the nature of its technical complexity but an idea that's valid nonetheless. I am therefore moving it to the Larger Suggestions Category. There, it will still allow folks to show support and help the Community Tech team communicate how the community perceives this need to the leadership at WMF. Thanks again. Regards, NRodriguez (WMF) (talk) 15:28, 28 January 2022 (UTC)[reply]

Voting