WEBVTT 00:00.000 --> 00:11.360 It's lovely to see you all here this morning, and hopefully you had a good evening last night 00:11.360 --> 00:16.080 and that there's no one who's feeling all that hungover after Saturday, it falls in 00:16.080 --> 00:17.800 that does tend to happen. 00:17.800 --> 00:25.800 This is just, everyone here know what alt text is, so it's alternative text, okay, you 00:25.800 --> 00:27.640 get that a little bit. 00:27.640 --> 00:33.760 So my favorite, I'm an open standard in practice lead as the team have said, and I work 00:33.760 --> 00:39.320 at civic actions usually largely with the US government, so we work on a large government 00:39.320 --> 00:44.000 website using Drupal, I'm also a Drupal Core Accessibility Meatiner involved in digital 00:44.000 --> 00:45.480 sustainability there as well. 00:45.480 --> 00:53.320 So the topic is how bad is alt text, and also, you know, what can we do about it? 00:53.320 --> 00:58.760 So alt text is really important, it's almost like the most basic form of accessibility. 00:58.760 --> 01:05.000 It's, it doesn't image have a description for it that allows somebody who uses assistive 01:05.000 --> 01:09.960 technology to be able to interpret and get sense of whatever that image was. 01:09.960 --> 01:14.320 So anything that's non-textual, because text information is quite easy to be brought over 01:14.320 --> 01:18.960 into a screen reader, the text is being traditionally pretty simple. 01:19.200 --> 01:26.640 The WKG framework, it's, it's 1.1.1, so it's, yeah, the very very first one, and, and it's 01:26.640 --> 01:32.840 a, it's also useful, if you think about being, accessibility is what people with disabilities, 01:32.840 --> 01:38.000 but it's also for people who have temporary and situational disabilities, so if you go 01:38.000 --> 01:40.840 outside in a bike day and you're trying to use your phone to try and figure out where you're 01:40.840 --> 01:46.160 going, if there's enough contrast in the screen, that will affect you also, you know, sometimes 01:46.240 --> 01:49.520 there's allergy seasons, so game, there's temporary in situation of disabilities that 01:49.520 --> 01:54.000 affect us all. It's not just people with permanent disabilities, but it's also looking 01:54.000 --> 02:01.840 to head to our technology, so Google Home, you know, Siri, are internet enabled, fidges and 02:01.840 --> 02:06.720 stoves, like it's, it's, we're, we're living in a world where so much more machine-readable 02:06.720 --> 02:11.200 content is interacting with, with web content, and, and the more we can make that 02:11.200 --> 02:14.400 semantically organized and, and, and, and, and transparent, the better it will be. 02:17.200 --> 02:23.760 So, how many people have used tools like, um, say, web aims, uh, wave toolbar, 02:25.120 --> 02:29.600 accessibility insights, is that an, it's another automated tool, how about site improve? 02:30.800 --> 02:34.320 Any, sites use site improvements, it's, it's a, it's a large sort of site scanning tool. 02:35.040 --> 02:39.440 There's a whole bunch of them out there, and they're, they're, um, very useful for a lot of 02:39.440 --> 02:44.880 different things. They, they generally prove, they don't prove accessibility. They prove that there's, 02:44.880 --> 02:49.520 there, their, their tests that are written, that can be evaluated, and they, they, they test, 02:50.480 --> 02:55.440 um, how, um, yeah, they, they, they, they, they, they test for it for, for a particular condition, 02:56.000 --> 03:00.640 um, but they cannot, they can't tell you if your site is accessible. They evaluate for about a third of 03:00.640 --> 03:04.240 what, what, what, the web, the web content accessibility guidelines are, so it's really useful, 03:05.120 --> 03:10.720 to have these automated tests, but they are, they don't allow you to go off and, and, and have a, 03:11.840 --> 03:17.120 determine without a site is accessible or not. Uh, Google Lighthouse is one of the big ones, um, 03:17.120 --> 03:21.360 that, and there's another tool called accessibility insights, also very useful. Um, 03:21.360 --> 03:24.800 one of them is done by Microsoft that's accessibility insights that they contribute back to the 03:24.800 --> 03:31.040 X-coil library, um, but Google little, it's a little startup, um, they don't contribute back to X-coil. 03:31.120 --> 03:36.800 So, um, yeah, we, it's a, the block, people do use these tools, um, and we need to try and 03:36.800 --> 03:40.880 invest in our open source libraries to do this and access it as a tool that's made, uh, this built 03:40.880 --> 03:48.640 by DQ, maintained by DQ, um, in the Netherlands largely by local fires. Um, and, uh, yeah, but 03:48.640 --> 03:54.320 all these tools, they, they evaluate the test that they evaluate for is, is all text there or not. 03:55.520 --> 03:59.520 It doesn't test the quality of the all text, so if you've got a value in there, 04:00.160 --> 04:08.400 it's tree or no or to be determined or quote, quote, it, they can't tell that and they don't give 04:08.400 --> 04:15.280 you a response to or falls for that. Um, so a lot of, a lot of tests are missing the quality 04:15.280 --> 04:20.720 evaluation for the all text. Um, and there's not even necessarily a good framework to determine 04:20.720 --> 04:26.640 how long should that text be? Um, NASA did a really interesting set of, uh, of all text well 04:26.640 --> 04:32.960 back on, uh, Hubble images where they, they took and they described these Hubble galaxy photos 04:32.960 --> 04:38.320 in intense detail, like there was a paragraph or two of all text in there. It was amazing. 04:38.320 --> 04:43.600 It was a really showcasing example of how a government site can try and make this content 04:43.600 --> 04:49.120 accessible to everyone. Do you want every image to be described in that much detail? I would think 04:49.120 --> 04:54.160 not because a lot of the images are not that interesting. Uh, so, you know, for any of us, like they're 04:54.960 --> 04:59.680 largely just slightly additional content. Um, there should be some information there, but, but it may 04:59.680 --> 05:05.680 depend on your user and it may depend on the context. And so it's very much depending on the author 05:05.680 --> 05:12.720 and the, uh, the effort to try and pull together. Yeah, what is relevant for the page and the content 05:12.720 --> 05:17.840 is there for it because because, because authors, authors don't add images or should not be adding 05:17.840 --> 05:22.960 images just because they need an image. They should be adding an image because it adds something to 05:23.040 --> 05:29.200 the value of the page. The user will have a better experience if they put an image there, which helps 05:29.200 --> 05:34.240 to enhance the content. So, so that's part of the challenge as well is that you're, you're really 05:34.240 --> 05:40.800 it's not just about describing an image. It's describing an image in context. So, um, so that's, that's, 05:40.800 --> 05:45.920 that's, that's, and, and to help them to convey the authors' intent. Um, because you're not just dealing 05:45.920 --> 05:51.840 with, with what is this image on its own. It's, what is the image in, in alignment with the text that 05:51.920 --> 05:58.000 is beside, and what is the message that the author is trying to convey? Um, so, anyways, there's 05:58.000 --> 06:02.480 lots of challenges with this. One of the other problems with, with these automated, automated tools 06:02.480 --> 06:09.760 is that most of them, um, giving it a scan on a page by page level. So, um, that's useful, um, 06:09.760 --> 06:14.880 if you're valuing a page, but often there's pages that go across, more than like there's problems 06:14.880 --> 06:18.640 that should, sure, across multiple pages. And how do you find out, when there's an image that's 06:18.640 --> 06:24.800 used in dozens of places on the site, where, where, well, it should have good alt text, um, or maybe 06:24.800 --> 06:28.560 it, I mean, maybe it should have the same alt text, maybe it shouldn't. Um, there may be cases where 06:28.560 --> 06:32.240 you have the same image, that's being used in a different context, and you want, actually, 06:32.240 --> 06:36.400 want to have a different image for that. So, um, that being able to see that and compare that, 06:36.400 --> 06:40.240 something that was really quite useful, and something that, that was missing from the set of tools. 06:41.280 --> 06:46.960 Um, and I wanted to also try to find ways to, um, to help authors find bad content and make 06:47.040 --> 06:52.800 it visible before it, before the public complained about it. Um, because that's part of the 06:52.800 --> 06:58.000 challenge with accessibility is that it is invisible for most users. If you don't know the tools to 06:58.000 --> 07:04.320 go off and expose the problems, you don't see it. So, how do you try and take the, um, how do you 07:04.320 --> 07:10.720 take that, that, that, um, how do you make these these problems visible and included as part of a 07:10.720 --> 07:15.440 definition of quality? Um, Google Lighthouse is really quite useful for that because it actually says, 07:15.520 --> 07:19.440 okay, you can, you can evaluate a page, it gives you a number of people can aim for that. This is a 07:19.440 --> 07:24.560 definition of quality is having a hundred on all of your, your Google Lighthouse scores. Like, that's, 07:24.560 --> 07:29.440 that's a useful goal for every page to have. Most pages do not comply to that, and it's useful 07:29.440 --> 07:32.960 for accessibility and performance and sustainability and a bunch of, it's a, it's a cute, 07:32.960 --> 07:37.840 it's a cute, it's a cute, it's a bunch of other stuff, but, um, but I, but again, we need to have a 07:37.840 --> 07:44.080 goal of what it is, and we haven't really had that with all text. So, yeah, if I were you 07:44.080 --> 07:51.200 repeated a bunch of this, um, and, uh, see what happened to sit here. Um, organizational compliance is also 07:51.200 --> 07:56.400 really important for many, many places. Um, so how many, how many people here are thinking about the, 07:57.680 --> 08:04.160 um, either the web accessibility directive or the, uh, European accessibility act. Okay, so this is, 08:04.160 --> 08:09.440 these are big issues. Like our, our industry is going to be really tested in junior July when this 08:09.440 --> 08:13.600 comes into effect, uh, with the European accessibility act. On the government side, it's already, 08:13.600 --> 08:18.720 in effect. Um, but it is, it'll be interesting to see how regulations change and how each country 08:18.720 --> 08:24.800 decides to apply the European accessibility act and how any international efforts tend to try and 08:24.800 --> 08:33.280 coordinate for that different set of regulations that are coming in for that. Um, and, um, the, the, 08:33.280 --> 08:36.880 uh, so also you know, how, how do you, how do you try and, if, if you have organizations, how many 08:36.880 --> 08:41.120 people who work for organizations here that have, um, websites that have multiple authors involved? 08:42.320 --> 08:46.080 Okay, Colleen, you've got a problem like, how do you try and maintain that consistency? 08:46.080 --> 08:51.440 Many of authors who are having all different levels of awareness around our own accessibility 08:51.440 --> 08:56.640 and that they, they don't necessarily have the ability to, um, to take the same training to 08:56.640 --> 09:02.240 care to write with as much detail on something like all text. So, so what I, I wanted to do is to 09:02.240 --> 09:07.120 have some way to do, to have a comparison for this. So, like a year or two ago, I did some work with 09:07.120 --> 09:12.480 Screening Frog, Screening Frog is an interesting NL, NL, analytics tool, Scamaging Webfoot, 09:12.480 --> 09:16.080 size for all kinds of information. I realized I could pull the all text out of that and create 09:16.080 --> 09:22.320 a list of the spreadsheets, do some analysis to say, uh, what, what kind of, um, data do I have? 09:22.320 --> 09:28.880 How can I see the, the alt text just as the, the image, the alt text and, uh, and, you know, 09:28.960 --> 09:33.600 yeah, any other title text or whatnot, pull that together into a spreadsheet to have it as a 09:33.600 --> 09:38.720 quick, simple analysis to compare for, for, for the, for our users, uh, or for authors, 09:38.720 --> 09:43.200 rather, to to evaluate and see how, how bad the alt text is. And just to get you curious, the 09:43.200 --> 09:47.680 alt text is bad on most websites. There are very few websites. Unless you're a specifically 09:47.680 --> 09:53.120 an accessibility oriented organization, it is bad. And even in those accessibility organizations, 09:53.120 --> 09:58.320 there are places where the, the alt text is missed or is just not done like it's, it is 09:58.400 --> 10:03.840 consistently a problem in terms of quality. Um, the other thing is, is that there's, uh, um, 10:03.840 --> 10:10.800 there's been changes in HTML, uh, any in here, know about long desk. So, so yeah, long desk is a, 10:12.080 --> 10:16.960 a fascinating debate if you want to go through the HTTP, um, mailing us and sort of see all the 10:16.960 --> 10:22.000 debates, but how do you deal with complex images? Um, alt text is really designed for, for short 10:22.000 --> 10:27.760 pieces of text, whether it's 180 characters or whether it's, you know, a paragraph or two, like, 10:27.840 --> 10:34.000 it's not designed for, um, a full page infographic with many different dimensions to it. So there 10:34.000 --> 10:40.240 are other ways to do that, but how do you evaluate whether an image has, um, has longer, uh, longer 10:40.240 --> 10:45.760 text associated with it? And long desk was an effort to do this in HTML, but has become depreciated, 10:46.640 --> 10:52.640 but there are still pages, even on the WC3 that uses outdated long desk format. How do you find these? 10:53.280 --> 10:58.000 Um, because the WC3C should not be using long desk if they're saying this is no longer a valid 10:58.000 --> 11:02.640 style at standard. They should be updating your information to maintain this. Um, alt wanted to 11:02.640 --> 11:06.480 have an updated open source tool, because I did, I was just annoyed by having to pay the, 11:06.480 --> 11:10.880 the licenses for, um, for a schooling frog, because I didn't have to use it that often, 11:10.880 --> 11:14.160 but everyone's the well, I wanted to use it and was like, ah, damn, I have to go and use this 11:14.160 --> 11:19.760 stupid thing and scan the whole website all over. Um, so I decided to create a new one. Uh, 11:19.760 --> 11:27.360 and I, F, ChatGBT, as you do, what's the best library to work with? Um, I'm a PHP guy. So I've got, uh, 11:27.360 --> 11:32.800 PHP was not the first language that they gave up for a large site calling, surprise surprise. Um, 11:32.800 --> 11:38.320 so, and that my old PHP guy, because I, you know, I've been, I've owned a business for 21 11:38.320 --> 11:43.280 years and now doing work, that's, that's tied to more policy-oriented stuff. So I'm not actually 11:43.280 --> 11:49.120 in a regular day, a programmer, even in Drupal. Um, but, uh, but yeah, Python was a choice. 11:49.120 --> 11:54.560 I don't know Python, but I do know ChatGBT, and so I was able to go off and pull together a tool, 11:55.280 --> 12:02.160 that was able to allow me to start scanning the websites and and evaluate this and pull information. 12:02.160 --> 12:08.000 It was fascinating. How many people have done programming with ChatGBT or other LLMs to try and 12:08.000 --> 12:12.640 pull things together? It is, it is fascinating. I mean, it's, it's, it's some of it's really terrible, 12:12.640 --> 12:18.160 and some of it's really good, and you just kind of, you have to, if like, working with a brilliant 12:18.240 --> 12:23.680 kindergartner, who will get the answer right, a third of the time, and now we're, we're, 12:23.680 --> 12:26.640 we're, you're often, and you get a really interesting script, and then you try and make a 12:26.640 --> 12:31.360 modification to it, and they, they delete half of it, and it's like, it's, it's, it's both 12:31.360 --> 12:38.480 infuriating and unspiring. How much can come, come out of these tools? Um, and, uh, the, 12:38.480 --> 12:44.400 yeah, math, just, it's all just math and statistics, but wow, um, it is amazing. Um, 12:44.480 --> 12:48.720 so I was able to go off and do this, and probably that was by going in and asking for specific 12:48.720 --> 12:54.800 functions and having it rewritten and doing many, many iterations over time. And so the script 12:54.800 --> 12:59.920 goes through, and I can ask it to go into and to call a call a CSV file, pick out a site, 12:59.920 --> 13:06.960 an XML site map, I can, um, look at at different other sets of logic to, to produce a CSV file, 13:06.960 --> 13:13.920 that then I can upload into Google sheet, or whatever else, um, and then then help share that with 13:14.000 --> 13:19.120 people, because, um, there are interesting ways to go off and pull other sets of data out, 13:19.120 --> 13:23.520 like Jason's another great way to go off into evaluate data. But if you're looking at sharing 13:23.520 --> 13:29.200 information with authors, a spreadsheet is the easiest way for them to see the information 13:29.200 --> 13:32.880 and to evaluate the information. So it needed to be something that was visible and easily share 13:32.880 --> 13:38.480 of all with people. Um, so I got some community feedback, because I've shared this as an open 13:38.480 --> 13:42.320 source project, and it was interesting to get that, that, that, that, that, that iterative feedback 13:42.320 --> 13:48.240 from them. Um, and yeah, as many of you will be asking, hey, I would be really cool for this. Yes, 13:48.240 --> 13:54.960 it is. Um, that is tricky too. Um, so, um, I've also enhanced the script to include some 13:54.960 --> 14:01.440 other things, um, because, um, each to know is a very powerful language and very complicated, 14:02.000 --> 14:06.000 um, and there's lots of ways to extend it to one of them is the, uh, accessible, which, uh, internet 14:06.160 --> 14:13.120 applications, which allows you to define, um, other sets of related content with that. So I've tried 14:13.120 --> 14:18.880 to go off into pull that information in, so we've got got some analysis there as well. Um, I was 14:18.880 --> 14:24.000 also able to provide some simple, um, text analysis for a quality, but, but then again, it's trying 14:24.000 --> 14:30.000 to, how do you get that standard of what is good and what it's not? So you can say, if it's less 14:30.000 --> 14:35.520 than 10 characters, then probably it's not sufficient all text. Um, you don't know that for sure, 14:35.520 --> 14:40.800 but you can say probably it is. If it's more than 350 characters, again, it's probably too long. 14:41.360 --> 14:47.040 Um, of course, uh, LLMs aren't really good at matching, um, characters. There are much better 14:47.040 --> 14:53.440 matching words than characters for some reason, but, um, at least in the current models. Um, so, 14:53.440 --> 14:58.560 uh, yeah, it was looking at ways to go up and that to evaluate the quality, um, I ran 14:58.560 --> 15:01.840 them through a language processor, which again, there's so many open source libraries that are 15:01.840 --> 15:06.640 available in Python. Things that I could just add on to say, let's evaluate the alt text. Is it 15:06.640 --> 15:10.160 is it written in plain language? Because it should be, like if the rest of your content is 15:10.160 --> 15:14.480 geared to be written in plain language, then your alt text should be as well. But the game is 15:14.480 --> 15:18.960 challenges with that, because sometimes you're describing technical content that summarizes 15:18.960 --> 15:23.440 in short and so, um, but, but again, we can evaluate to see if there's risks that, 15:23.520 --> 15:30.080 some of you can look at the code and try and evaluate that. Um, I want to provide some real 15:30.080 --> 15:36.160 time examples of that alt text. Um, these are from existing websites, both in Europe and the United 15:36.160 --> 15:43.440 States. So, um, and we can go to the slides you can access this, but alt equals quote quote is not 15:43.440 --> 15:47.200 effective alt text. This should never be in the back. It's of alt equals quote quote. So just, 15:47.200 --> 15:51.840 you know, or quote quote just to be clear, even if you wanted to be an alt text, putting quotes in, 15:52.480 --> 15:56.160 makes it invalid. That is, that will screw up a screen reader and that the screen reader will have 15:56.160 --> 16:02.080 it pronounced to them quote quote. Um, this is a classic one. Um, these two are like final names that 16:02.080 --> 16:06.240 dropped a J.P. off of them because that's, you know, somewhat nice, but it's a final name that somebody 16:06.240 --> 16:12.480 went off an uploaded. And it's like, is that, like, maybe this is relevant because the final 16:12.480 --> 16:18.960 name has some useful information in it, but most of the time it doesn't. Um, here's another one 16:19.040 --> 16:25.120 here. I can assume this means illustration. Okay, but again, how does that help me? There's 16:25.120 --> 16:30.880 an illustration, excellence. Like, if you, I have not, what I have not done, there's another approach 16:30.880 --> 16:35.920 to try and take these alt text and to try and convert them into images. Because you shouldn't 16:35.920 --> 16:40.960 do that, right? You should be able to take the, the, the alt text here and run it through an image 16:40.960 --> 16:45.600 process. They're going to say, what, what do I get? And how does it fit? How does it compare? I mean, 16:45.600 --> 16:50.160 if that would be a terrible waste of energy, um, because from the CPE energy conflict, like that, 16:50.160 --> 16:56.640 yeah, there would be a terrible waste of energy. Um, but, uh, he had no, also bad, um, photo of working 16:56.640 --> 17:01.920 group one. I mean, a little bit better, but, I mean, there's, there's a group of people around 17:01.920 --> 17:05.680 a table, no, four of them. It's a photo of a group of one, photo of working group two, photo 17:05.680 --> 17:10.320 of working group three. Like, that's not useful information. And then you see that and to 17:10.320 --> 17:14.640 have an online, when you expose that information, you can start to say, I think we can do better 17:14.640 --> 17:23.920 than this. Um, so yeah, share early, share often. Um, they'll be a link to the script, uh, in the, 17:23.920 --> 17:28.000 in the, in the slides, but also, um, I think it's a link to in the, the, the, the, the, the, the, the, 17:28.000 --> 17:34.000 the presentation as well. Um, so you can run this little Python script, um, and it will go through 17:34.000 --> 17:37.840 and pull this information. Um, there's also a reading file that's a little bit out of the date, 17:37.840 --> 17:42.720 but chat with GBT is quite useful for writing reading files, because people suck at it. Um, 17:42.800 --> 17:46.240 and especially when you update your code and you're trying to remember what, what is in it? Like, 17:46.880 --> 17:51.360 at least it's somewhat more useful to go off and have something that is a, yeah, it's better than me, 17:51.360 --> 17:58.320 right, uh, that kind of documentation. Um, but, uh, but yeah, please go off and we want to have 17:58.320 --> 18:01.840 contributions from people. We want to have a discussions around this. There are some other 18:01.840 --> 18:07.360 interesting tools that are looking at, uh, having this discussion as well. Um, one of them is, 18:07.520 --> 18:14.000 um, editorially, um, so editorially and, and Sally, these are two tools that are editorial tools. 18:14.720 --> 18:18.880 Um, it's a, there's a talk on a tag that's going to be later on today, and these are really 18:18.880 --> 18:23.360 both interesting tools around how to improve the offering experience, how do you make it easier 18:23.360 --> 18:27.840 for them to see where the problems are? Um, they being looking at struggling to how to, how to 18:27.840 --> 18:34.800 include, um, all text in that. Um, so talking with John Jamison and others, um, to try and, and pull 18:34.880 --> 18:41.120 together, um, some experiences they've had, um, and we need more help than work guidance in this, 18:41.120 --> 18:44.480 and we've been doing this all in English. Um, there's going to be different sets of rules for 18:44.480 --> 18:49.040 different countries. So the English, the English languages should not be assumed to be the default 18:49.040 --> 18:53.440 for everything. Like, I don't know what the best practice for a French or German all text is, 18:53.440 --> 18:59.280 especially with the length of German words. Like, sometimes, yeah, localization was really 18:59.360 --> 19:08.480 important, and that also comes down to all texts. Um, so, um, feedbacks really important for 19:08.480 --> 19:12.160 trying to go off and evaluate this. Authors need to have that feedback. There needs to be a cycle. 19:13.440 --> 19:19.120 Um, open source tools are really helpful to try and and build that collaboration. Um, and, and it's 19:19.120 --> 19:25.600 it's interesting how, I don't think that that there's a, um, that this type of tool has existed 19:25.600 --> 19:31.360 before, um, at least in my search and effort to do that. Then, of course, I came here and was talking 19:31.360 --> 19:36.320 with TBO yesterday, and there's apparently already a tool that TBO was created that does much of the 19:36.320 --> 19:40.960 same thing. So it's sometimes amazing how this type of stuff can happen, um, but we need to 19:40.960 --> 19:45.600 learn from each other's artwork and share that and be able to, to push it further to see how 19:45.600 --> 19:51.760 we can make this be something that is, is beneficial, beneficial for all of us. Um, and, you know, 19:51.760 --> 19:56.800 accessibility is something that, that, that is everyone's responsibility. So, uh, we need to be able 19:56.800 --> 20:02.000 to help and look at all of our teams, whether it's the content teams or the developers, the designers, 20:02.000 --> 20:06.320 uh, there's a whole sort of teams-based approach, and I could happily give another talk about 20:06.320 --> 20:10.560 accessibility roles within development teams, but I'm not going to give that now. Don't bring. 20:12.320 --> 20:17.200 Um, so yes, I did incorporate AI into this as well because I'm YouTube. Um, 20:18.160 --> 20:22.960 and, uh, it was really quite interesting to see that comparison of, uh, when you actually take the 20:22.960 --> 20:28.960 text that authors are given versus the, the, the, all texts that, that, um, a good LLM can generate. 20:30.240 --> 20:35.920 Um, and it's, it's quite fascinating to see how, how good it can generate the content, like the, 20:35.920 --> 20:40.640 the, it, it, it, it's providing much, for so much more content than that I would ever write for all 20:40.640 --> 20:44.240 texts. Um, but then I think, well, what is the right link? How do you want to trust 20:44.240 --> 20:48.560 structured? And what are, what is the author focusing on? Um, what is, what are the pieces you want 20:48.560 --> 20:52.480 to highlight when you want to drop out? Um, and how do you build that so that you're actually 20:52.480 --> 20:56.560 getting something that is a useful tool? Um, and again, as more, there's people who've explored this 20:56.560 --> 21:01.040 as well, but how do you provide a prompt to make sure that you're getting the right kind of feedback 21:01.040 --> 21:06.560 from people? Um, and yes, so the tool that I built does go through and, and provide a couple of 21:06.560 --> 21:12.000 different solutions, um, and allow you to evaluate them to say, like, well, at least if an author 21:12.080 --> 21:17.920 looks at the, the errors, or looks at the suggestion, they can edit it and evaluate it. The author 21:17.920 --> 21:21.520 still does need to be in the loop, but it's quite useful to go off and to see what the tools 21:21.520 --> 21:27.120 and technical, what, what does a bot suggest? And how do you try and take that, that suggestion, 21:27.120 --> 21:33.440 and move, move from that to something that actually reflects the needs of the author. Um, 21:33.440 --> 21:38.080 and, uh, yeah, it's also still not sure where we're still struggling with this in terms of 21:38.080 --> 21:43.040 what, uh, LLMs are possible. In terms of attribution, um, because that's another piece of this, 21:43.040 --> 21:47.040 how do we try to make sure if you're using an LLM to generate all text? How do you, uh, 21:47.040 --> 21:52.400 what all attribution is required? How do you credit that, uh, whatever the source of information is, 21:52.400 --> 21:57.840 is it necessary to credit whoever is behind that for the creation of the code? Um, and that's 21:57.840 --> 22:02.640 something I still don't know. Um, and yeah, there's another text that's, that's in the same repository 22:02.720 --> 22:08.320 that that allows you to, to hit a number of different, uh, uh, open, or, uh, different language models 22:08.320 --> 22:18.160 and evaluate them, whether it's, uh, Anthropoc, uh, or, um, Chad GBT, or, um, Ollama, and there's a couple 22:18.160 --> 22:25.040 of other ones as well. Um, so yeah, evaluate your websites with, with these tools, reflect on the 22:25.040 --> 22:29.520 meaning of all texts, think about how you can go off and make it better. Joining community efforts 22:29.600 --> 22:34.320 to try and, and look at this, these types of tools and see what kind of guidance we can give, 22:34.320 --> 22:39.200 give authors, help, because the, the efforts that we've made to produce good all texts in the 22:39.200 --> 22:43.840 last 30 years have not worked. You know, there's lots of pages and how to write good all text, 22:43.840 --> 22:47.840 but they don't, it doesn't matter how many good pages there are, we don't have good all text. 22:47.840 --> 22:51.760 Um, so how do we try and find a way to make this possible? Also check out the, uh, 22:51.760 --> 22:56.560 Drupal CMS module. There's, as Drupal CMS is a new version of Drupal that's come out 22:56.640 --> 23:00.880 that has any, I function built into it. And the AI tool within it is quite useful. 23:02.000 --> 23:06.640 Um, you can, once, once you've got it set up, you can upload an image and immediately 23:06.640 --> 23:11.680 it gives you an opportunity to go off and to fill in the, the, the text behind that, um, and edit the 23:11.680 --> 23:16.640 text before you hit submit. So it's something that's fairly penless for the author. Um, and that's it. 23:16.640 --> 23:20.160 That's all, all the, all the, all the stuff I have. 23:20.160 --> 23:26.640 Yeah. And hopefully reach out. 23:30.240 --> 23:35.760 Yes. What would be appropriate all text for a logo? Is that something you should just write 23:35.760 --> 23:40.160 as it is? So do we just say, it's a, well, you said it's a fat example of what would be good 23:40.960 --> 23:46.000 text for that. Apparently depends what it does. If it's on it, but it's a logo on its own, 23:46.000 --> 23:51.280 you might want to describe the logo. Um, you, uh, especially if it's, is a profile, if you're, 23:51.280 --> 23:55.680 if it's a page where you're talking about the logo, um, if it's, if it's a, if it's a link, 23:55.680 --> 23:59.840 like on the home page, you probably want to say, you know, more that, that, that, that, that, that, 23:59.840 --> 24:04.160 that you want to talk about where, what the link is going to, and what is it doing? Because, 24:04.160 --> 24:08.160 uh, people don't want to hear, you know, that a long description of what your logo is every time 24:08.160 --> 24:14.960 they load the page, right? Good. Um, about any concrete guidelines on how long it will 24:15.040 --> 24:19.840 text can be because some years ago, I was told, um, I'll text should be too long, but it's the 24:19.840 --> 24:24.960 reasons and stuff. Well, SEO should never be the, the reason why you're, you're doing anything for 24:24.960 --> 24:28.880 all texts. Like, it ultimately, that's not the user that most of us should be aiming for, um, because 24:28.880 --> 24:33.760 it's, we're humans and humans, you know, yeah, but, but people do use all texts for SEO. Um, 24:33.760 --> 24:38.720 but, uh, Eric Egerts, who's from Germany, who's a wonderful accessibility expert that I 24:38.720 --> 24:43.040 looked you all the time, he's done some extensive testing in popular all, uh, screen leaders, 24:43.120 --> 24:48.560 and he's like, there is no practical limits on how long the all texts can be. So you can make 24:48.560 --> 24:53.600 it as bloody long as you want. Is that a good idea or not? Like, well, that's, that's, you know, 24:53.600 --> 24:57.280 but you're not technically limited. It's, it's more a question of how practically long should 24:57.600 --> 25:05.360 be. Yes. Similar question, I've been accused of writing to long out texts because of the balance 25:05.360 --> 25:13.600 between, uh, out texts to, like, describing and inclusion, so describing, like, the physical nature 25:13.600 --> 25:20.480 of people included in things, to be more inclusive. Yeah. Is there good guidance on how to do that 25:20.560 --> 25:28.400 again without, so you're describing everything? No, and it's also, that's also a very political 25:28.400 --> 25:32.960 thing where it's in some countries now, it's more of a dicey issue than others. Um, but I would say 25:32.960 --> 25:40.240 that, uh, probably, what I'm hopeful for, there's a drive to go off and embed, um, the, this 25:40.240 --> 25:44.960 images, the, all texts are image reading in, in your in browsers, where that's Chrome or Firefox, 25:44.960 --> 25:49.440 something that's a Firefox talk on this later on today, somewhere. Um, that's a really interesting 25:49.440 --> 25:53.440 initiative, but we don't actually want that to be done as the, the default. We want to have, 25:53.440 --> 25:57.040 we want to go have people choose it. If you want, if an author wants to have a verbose 25:57.040 --> 26:01.680 all texts or to query an image, we want that, because I think, like, default, you want to have 26:01.680 --> 26:06.560 it as simple as possible and allow the user to expend it to something that allows them to see 26:06.560 --> 26:11.120 the full context and what's possible as required, but not on the, they have the choice. They should 26:11.120 --> 26:14.320 have the choice. Thank you, Mike. All right.