Wikidata:Property proposal/alt text

alt text edit

Originally proposed at Wikidata:Property proposal/Sister projects

   Done: alt text (P11265) (Talk and documentation)
DescriptionAlternative text used when an image can't be rendered or when accessed from assistive technology such as a screen reader.
Representsalt text (Q60844786)
Data typeMonolingual text
DomainCommons image
Example 1File:Ansel_Adams_and_camera.jpg » "Black and white photograph of a man with a tripod camera and pine trees in the background."
Example 2File:Apollo_11_Crew.jpg » "Three men in white spacesuits without helmets in front of an image of the Moon."
Example 3File:Panthera_tigris_altaica_13_-_Buffalo_Zoo.jpg » "Tiger cub in front of an adult tiger in the snow."
Planned useUsage in various tools as well as on Wikimedia Commons and Wikipedia.
See alsoformer proposal for Wikidata

Motivation edit

This will allow editor to specify an alt-text(alternative text which describes the visual features of an image) for images on Wikimedia Commons which can then be reused by various projects and tools to set an alt-attribute. See the tickets below and the previous proposal.

See Phab:T166094 and Phab:T213585 Abbe98 (talk) 20:09, 20 February 2021 (UTC)[reply]

Discussion edit

  •   Support Good idea! Thanks for explaining the process. Jane023 (talk) 20:16, 20 February 2021 (UTC)[reply]
  •   Support --Lectrician1 (talk) 01:29, 21 February 2021 (UTC)[reply]
  •   Support   Don't ping me NMaia (talk) 10:36, 21 February 2021 (UTC)[reply]
  • Strongly oppose per discussion in the previous proposal, and Phab. tickets, since when nothing of significance has changed. As I said in the first Phab. ticket - and as I advised the proposer, in the second - "Please do not progress this proposal without first consulting with screen reader users and/or accessibility professionals." It appears this had not been done. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:16, 21 February 2021 (UTC).[reply]
    Please provide some constructive feedback rather than repeating a statement that questions others expertise. I have worked professionally with web accessibility. Also note that this is not about implementation of alt-text but about a way to store generic alt-texts. Abbe98 (talk) 15:04, 2 March 2021 (UTC)[reply]
    Oh FFS. I have provided plenty of constructive feedback on past iterations of this benighted proposal. I am sick of people trying to wear down sensible and reasoned opposition including that by colleagues who are blind and use assistive technologies like this. I too have "worked professionally with web accessibility" - that doesn't make me an "accessibility professional" and nor does it make you one. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:57, 2 March 2021 (UTC)[reply]
  •   Comment From the domain in the proposal, I note this is limited to Commons (not for use on Wikidata itself). So from Wikidata's POV, there isn't really any reason not to create this. It seems that Monolingual text properties are now possible on Commons [1], so technically, this could be implemented. Maybe the property description could recommend reading some help text that explains how to write alt texts. --- Jura 11:31, 21 February 2021 (UTC)[reply]
  •   Support This really is common sense but in any case I asked some people with relevant expertise to comment. (Andy you might want to explain or link to your objection to make it easier for them to assess it.) I'll also note that there are various places where we cannot add alt text currently (category pages, image description pages, generated reuse snippet etc) and this proposal would help with that; the benefits would not be limited to fallback alt text in articles. Also, having a large corpus of well-described images might have value of its own for developers of assistive technology. --Tgr (talk) 15:19, 21 February 2021 (UTC)[reply]
  •   Comment I imagine the eventual data type of this property would be multilingual text, but for the time being it should be created as monolingual, as done for other properties. --Tgr (talk) 15:26, 21 February 2021 (UTC)[reply]
  • "planned use = Usage in various tools as well as on Wikimedia Commons and Wikipedia." (my emphasis) Wikipedia is an encyclopedia and only includes an image for the information that the image provides to the reader that is relevant to the article . It's not there to make the page look pretty. On Wikipedia, images are almost invariably accompanied by a caption that supplies additional information not apparent from the image. It is the combination of alt text and caption that makes up the alternative text for the image. That means that the alt text is context dependant and cannot be prescribed from a central location. If you place File:Washington and Lafayette at Valley Forge.jpg in an article about en:Valley Forge, it doesn't matter that there are two officers with red capes and five soldiers, but it's important to say that it's winter. If you put it into an article about en:George Washington, it doesn't matter what colour the horses are, but it probably does if it's in the article about en:Nelson (horse) (he's the one on the left). If you want to annoy screen reader users, put File:Ansel Adams and camera.jpg into the en:Ansel Adams article with the alt text suggested as an example. Try explaining to them what relevance "a pine tree in the background" has to his article. Was he notable for standing in front of pine trees? By the way, the suggestion here would still be an improvement over the abomination actually used in his article ("the open shirt collar is spread over the lapel of his jacket" - suitable for a fashion magazine, not for an encyclopedia). --RexxS (talk) 02:05, 23 February 2021 (UTC)[reply]
This is all very true and I believe it important to note that this proposal is only a proposal to have a way of storing generic alt-texts. Wikipedia might choose to somehow use this and it might also choose not to. In my opinion Wikipedia editors and should always be in control of the alt-texts in articles. Abbe98 (talk) 15:00, 2 March 2021 (UTC)[reply]
  • Oppose as a screen reader user – nothing's changed since last time. Graham87 (talk) 06:16, 23 February 2021 (UTC)[reply]
  •   Oppose on philosophic grounds, and per views of previous opposers. I don't like Wikidata increasingly building itself as "the one true God" of all that is Wikimedia. Context matters in alt text as well as captions (as described at en:Wikipedia:Manual of Style/Accessibility/Alternative text for images). What happens on Wikidata (or Commons) need not necessarily be reflected in every source that uses Wikidata data (or Commons media). Poorly curated infoboxes will accept whatever is pumped into them from Wikidata, regardless of relevance. The existing media legend (P2096) for example often results in a redundant, unhelpful and/or arbitrary foreign language caption in Wikidata Infobox (at least when implemented on Commons), unless one adds an arbitrary legend in their preferred language. No matter how well the intentions are, the more centralized and prescribed data presentation becomes, the harder it becomes to maintain local control on individual articles, infoboxes, and whatever third-party sources draw from Wikidata. Someone will probably eventually make a bot that registers all incongruences between local and Wikidata values as "errors", giving some other bot operators incentive to wrongly "fix" them. -Animalparty (talk) 01:41, 24 February 2021 (UTC)[reply]
  •   Support Very good! This will facilitate translations, make pictures better searchable and provide material for the developers of image description AI. Most of all it will make it more effective to provide images with alt texts. Regarding the context issue: In my opinion, the image related text form that should deliver the context is the CAPTION. Context is necessary for ALL users. So it should be readable for everybody, but alt text is only readable for screenreaders or when image display is disabled by the browser. Alt text should only replace the VISUAL CONTENT of the image. It should tell those who cannot see the image what people who can SEE. In Wikipedia every image has FIVE image related text forms: Name, caption, alt text, image description, article. The caption changes from article to article so it should give the context. The longer image description does not change but since its longer it can be sensitive to different possible contexts. The alt text should change as little as the image's file name. It's not necessary and would be even redundant if the alt text would repeat the caption. KH32 (talk)16:39, 25 February 2021 (UTC)[reply]
    • @KH32: A user who is blind and who uses assistive technology and relies on alt text attributes commented against the previous proposal, explicitly on the basis of context; and reiterates their opposition here. Can you explain the basis on which you have a better understanding of this matter than they do? And can you explain how this will make images "better searchable" than the proper use of "depicts" statements will? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:57, 2 March 2021 (UTC)[reply]
    • @Pigsonthewing:I am speaking for the group of blind and sighted people of our Photo Studio for Blind Photographers in Berlin. We are working with picture descriptions since 2011. But that doesn't mean that we have automatically the better understanding in this particular question. The reasons why we disagree in the question of context in alt texts are just as I tried to state them briefly. Regarding your second question what do you mean by "depicts" statement? KH32 (talk)18:59, 02 March 2021 (UTC)[reply]
    • @Pigsonthewing: Ah that alt texts = alt tags make pictures better searchable is because "Google uses alt text along with computer vision algorithms and the contents of the page to understand the subject matter of the image."(Google) KH32 (talk)19:17, 02 March 2021 (UTC)[reply]
  •   Oppose if this should be implemented, it shouldn't be implemented as a property, but using the empty "descriptions" array, see phab:T166094.  – The preceding unsigned comment was added by Multichill (talk • contribs) at 18:56, 3 March 2021 (UTC).[reply]
  •   Support to provide a default, and as alt-text for the Commons information page; and for when the image appears without context, for example on a wikidata page. Where the image appears elsewhere, make sure there is provision to override the alt-text locally. Jheald (talk) 22:51, 3 March 2021 (UTC)[reply]
  •   Comment On the Phabricator task, User:FRomeo (WMF) says: We're thinking of organizing a workshop on this topic, with accessibility consultants, people with lived experience, the WikiBlind user group, and anyone else who is interested. - which sounds to me like the best way forward for this proposal. --Tgr (talk) 10:35, 5 March 2021 (UTC)[reply]
  •   Support (would also support using the "descriptions" per Multichill, but we need it so bad I won't say it's a blocker) Ainali (talk) 12:27, 1 July 2022 (UTC)[reply]
  •   Support It is only the description of the image for those who cannot see it and use screen readers to read the text on the internet, and not, as some opponents say, text related with context, opinion or subjective things. Even if it's not really used anywhere else at the moment, it would be a very important step forward for blind and visually impaired people, not to mention that it will eventually allow images to be more easily classified according to their content. Şÿℵדαχ₮ɘɼɾ๏ʁ 22:35, 9 August 2022 (UTC)
  • There is consensus on Commons to support offering a default alt text for files in different languages that can be overridden by users on local wikis if desired. This should probably be handled through SDC in one way or another. If a property is needed for that, I   Support the creation of one, although it has been mentioned that something akin to the multi-lingual "captions" might be a better solution. --El Grafo (talk) 14:08, 31 August 2022 (UTC)[reply]
  • @Multichill, KH32, Animalparty, Graham87, RexxS, Volker E. (WMF):@Tgr, Jura1, Pigsonthewing: Given the clear support for the property created as alt text (P11265) ChristianKl17:36, 4 December 2022 (UTC)[reply]