20

We had a request a couple of years ago to replace "enter image decription here" with some sort of generic text. I think this is a bad idea, as it doesn't add anything. Shog's comment on the proposal essentially articulates how I feel:

I would rather it changed to something like, "Author hates blind people".

So our only other option is to enter useful descriptions by hand.

This Google search shows which posts are still in need of help.

I will make a list of posts that still require fixing, and we can remove them from the list as we go. Feel free to add any that you find yourself.

You don't have to write like Shakespeare or Nabokov, just something that makes it clear what the picture is of.

Obviously this is a big task, but each picture adequately described is a little victory!

If you're interested in tips on how best to fill out an alt text, this site has some ideas.

  • 1
    Thanks for doing this, Matt! – Kit Z. Fox Oct 14 '14 at 16:24
  • So, how long before Google updates its search and we can get a new batch to edit? Right now, the search link returns a bunch of questions that have already been edited. – Marthaª Oct 14 '14 at 18:41
  • I'll go through the search and find some more candidates. Thanks for you help, @Marthaª. – Matt E. Эллен Oct 14 '14 at 19:58
  • What to do about Google nGram plots? – 200_success Oct 17 '14 at 21:34
  • 1
    @200_success: I've been doing my best to describe them. "Phrase A (in red) has its first peak in 1850, then it gradually increases to its highest peak in 1900, after which it gradually declines to below its 1850 peak by 1990. Phrase B (in blue) first shows up in 1920 and lags behind Phrase A until around 1950. There are crossover points in 1960 and 1975, but after that, Phrase B starts outstripping Phrase A, and by 1990 it rises above Phrase A's highest peak." Or some such. – Marthaª Oct 18 '14 at 0:41
  • color is irrelevant. – SrJoven Oct 24 '14 at 17:12
  • 2
    Please look at this link regarding alt text. It appears I've been doing it ... overkill. The images should not have to be overly descriptive. In fact, the public facing written context of the post should be adequately describing the image anyway. If that happens, the correct answer for non-link images could be alt="", with the appropriate text visible. – SrJoven Oct 25 '14 at 2:48
  • @SrJoven colour might be irrelevant, I could imagine that severely short sighted people might utilize these screen readers too. But I'm not sure, never having seen these devices myself. No doubt that in the US these are common tools. "Overkill" is exactly wanted to say in my post, an overload of information and detail, which becomes almost counter productive. And then what about Ngrams? How would you describe similar charts where results spike suddenly or slowly curve upwards/downwards etc. – Mari-Lou A Oct 25 '14 at 12:18
  • @Mari-LouA There are at least two reasons alt attributes (pedants say tag is the wrong word, but it's easier to type) are useful: 1) when the image is unavailable. and 2) when the screen reader sees it and reads it aloud. For [ngram] charts, it's probably more important that the gist of the graph is in plain text (As you can see, foo has a lot more usage than bar) and the alt="Google ngram chart". – SrJoven Oct 25 '14 at 12:25
  • @Mari-LouA but the interesting thing is that if the image is already introduced in text as an ngram chart, the alt tag can be blank ("") per the link above. All the description is already in the visible text. – SrJoven Oct 25 '14 at 12:28
  • @SrJoven that depends, users will often say something similar to: see Ngrams chart, Ngram agrees with me, etc. – Mari-Lou A Oct 25 '14 at 12:31
  • @Mari-LouA Oh, I don't disagree with what is. I'm saying what should be done. That is to say, as ahem cough cough informative as the chart may be, it's probably not as important to casual observer the content as much as the implication. In other words: Yes, it may be interesting, but the relevant part is that the chart supports the answer. Not that the chart is the answer, which after all should be what's visible text. – SrJoven Oct 25 '14 at 18:57
  • @SrJoven: the problem is that two people can look at the same NGram chart and one of them can decide that it supports the argument, while the other one can decide that it doesn't support the argument. As sighted people, we can look at the chart and make our own conclusions about its validity, but if all a blind user hears is "this is my argument, and here's a chart to support it", then they have no way of knowing if the chart doesn't actually support anything. – Marthaª Oct 28 '14 at 17:07
  • @Marthaª Given that, a) link to the chart source and b) the picture content of NGram is still irrelevant to the context. Describing it (invisibly in alt) is only going to be confusing. – SrJoven Oct 28 '14 at 17:48
  • We just started a similar effort at Mi Yodeya. We use many fewer images, so we have a shorter hill to climb. Good luck! – Isaac Moses Jan 29 '15 at 22:03
3

I've made a Stack Exchange Data Explorer query to find posts that lack image descriptions. As of 2014-10-17, the query shows about 1500 posts that need work.

Please keep in mind that Data Explorer results are based on weekly extracts from the database, and therefore may not reflect recent changes.

4

To do:

(got to page 11 on the Google search)

Thanks to everyone for getting on this so quickly!

Done:

  • did "what does it mean" – Ludwik Oct 22 '14 at 9:16
  • Thanks, Ludwik! – Matt E. Эллен Oct 22 '14 at 9:49
  • 2
    Argh, I hate doing nGrams graphs! – Mitch Oct 24 '14 at 18:00
  • @Mitch: I'm right there with you. I find myself wondering if I can just summarily delete graphs that aren't absolutely vital to the question/answer. I'm also growing to detest screen captures and pictures of old newspapers and other images of text. :/ – Marthaª Oct 24 '14 at 21:39
  • @Marthaª wouldn't the solution be "newspaper clipping illustrating how (word) is relevant/related to answer" There has to be a summary of some sorts. – Mari-Lou A Oct 25 '14 at 19:03
  • What, that's it? – Mitch Oct 25 '14 at 20:44
3

The value of doing this is highlighted by today's Imgur downtime. We have questions like "A pronunciation question of slough" which make no sense without the embedded image.

This is a problem for sighted users today, and it's a problem for blind users all of the time.

  • Ah, so that's what's going on. I thought my new PC had some issues. – NVZ Mar 1 '17 at 9:43
  • Yeah, but what's the alternative? Type out the dozen+ "ough" words in the alt-text? That defeats the purpose of using an image in the first place. For that example, I think the best we can expect people to tag is "Series of words containing the string "ough" having very different pronunciations". – Dan Bron Mar 1 '17 at 14:08
  • 1
    @DanBron The unintended consequence of using an image is to make the site useleess for the blind. In this case, the image contained text. So yes the obvious alternative was to type out the text. – MetaEd Mar 1 '17 at 14:47
  • Maybe I'm missing something. How does a blind person read the alt-texts (or even anything on a computer screen)? Text-to-speech plugins? – NVZ Mar 2 '17 at 4:24
  • @NVZ Text to speech is most common. Some use a Braille module. – MetaEd Mar 2 '17 at 13:55
  • @DanBron what was the purpose of having it be an image in the first place? a little chunk of text formatted with markdown could have replaced it. – Hellion Mar 7 '17 at 20:30
  • @Hellion The original author of the image clearly created it so that all the "ough"s would line up vertically, making the cognitive dissonance in their different pronunciations even sharper. The OP who used it in the post did so because it offered a simple, direct, and low-effort way to illustrate the variability of English orthography (or pronunciation, depending on how you look at it). Asking that person to then type out the whole image defeats that entire purpose, and therefore makes barriers to entry higher, and therefore reduces the likelihood of getting (quality) posts. – Dan Bron Mar 7 '17 at 22:12
2

You don't have to write Shakespeare or Nabokov...

That's reassuring. So, it's not necessary to be one the greatest playwrights ever known to mankind or probably the most iconic 20th century novelist to fill in the image description. Good to know!

But according to an edit of yours, where the image is a photo I took of the Irish Daily Mail, the description has to be pretty detailed.

Issue of the Irish Daily Mail from Monday, June 30, 2014, folded in half. Banner image has three columns, first is a picture of a woman, second is the text "Miriam: I've gotten better looking as I get older see page nine", third is the text "only €1 a big read for a small price". The third's background is a baby blue circle. The whole banner image has a gradient background going from red to black. Below the banner image is a by line: Accidents, fires and breakdowns mean vehicles 'not fit for purpose'. The main headline is below the fold.]

I mean I love the description, don't get me wrong. It's beautifully crafted, I wouldn't be surprised if Google picks up on it, but it does seem a little excessive.

From now on are we expected to go into so much detail? For heaven's sake why?

  • 1
    People go into as much detail as they think necessary. – Matt E. Эллен Oct 20 '14 at 14:22
  • @MattЭллен Could you explain its relevance? I'm no computer geek, it must be obvious to programmers but on me it's lost. – Mari-Lou A Oct 20 '14 at 15:06
  • 1
    The description becomes what is known as the "alt text" of the image. This is useful for people using text only browsers and for people who use a screen reader. Mostly, it's to help blind people (they are the major group who use screen readers). A screen reader is a web browser that reads the contents of the screen out loud. – Matt E. Эллен Oct 20 '14 at 15:09
  • 3
    Oh. Presumably it's a machine that reads text out aloud. And if that's the case, surely the aim of the description is to pinpoint the focus? Headline title "bla bla" with a photo of a blonde middle-aged TV presenter called Miriam. The picture shows the reader that the attractive woman in question is not in her 30s (or 40s). Whereas saying the paper is folded in half, it has three columns, a blue circle, a gradient red to black, and quoting the second headline is, I think, distracting for a blind person, there's an overload of information which all appears to carry equal importance. – Mari-Lou A Oct 20 '14 at 15:31
  • This should be a separate question, not a question masquerading as an answer. – Marthaª Oct 20 '14 at 15:49
  • 4
    If the image can be described adequately by text, shouldn't it just be replaced altogether? – Barmar Oct 20 '14 at 15:58
  • 2
    @Marthaª I was asking for clarification, I wasn't aware that Matt Эллен was asking a question. My post is clearly asking why, I'm not acting sneeky in the slightest. Tsk! – Mari-Lou A Oct 20 '14 at 16:01
  • 1
    @Barmar I think images can save a great deal of time and space. Somehow when you actually see a ship sinking, it becomes more believable, more real. Likewise the image of a newspaper headline stops anyone from doubting that you invented, or incorrectly transcribed the whole thing. – Mari-Lou A Oct 20 '14 at 16:06
  • @Mari-LouA "Images can save a great deal of time" Not if we're doing descriptions that accurately cover what's in the image + uploading the image, compared to just writing the description. "a newspaper headline stops anyone from doubting that you invented, or incorrectly transcribed the whole thing". A citation can do that too. It's also worth remembering we live a Photoshop age. "I think images can save a great deal of time and space" Text typically requires smaller data storage, and therefore makes it cheaper for SE to provide. However I agree images make it more real. – Robin Hood Oct 22 '14 at 5:05
  • @Mari-LouA please feel free to edit any of the descriptions you want to improve. I'm far from perfect at describing things! – Matt E. Эллен Oct 22 '14 at 9:57
  • 1
    Alt text should be minimal (and I'm guilty of not following this. See my comment on the master post). Note that screen readers read aloud everything on the screen. If the relevance of the image is described in public facing text, the content of the image should not replicate that information. – SrJoven Oct 25 '14 at 11:46

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .