WEBVTT 00:00.000 --> 00:10.540 Okay, afternoon everybody, as this part of the slide says, my name is Steve and as this 00:10.540 --> 00:12.780 part of the slide says I'm a bit of a geek. 00:12.780 --> 00:16.540 If we've all turned up to the correct deaf loom, this is to talk about grammar. 00:16.540 --> 00:22.480 Yeah, I should probably tell you from my accent, I'm English, anyone else here from England? 00:22.480 --> 00:24.480 Welcome to Europe. 00:24.480 --> 00:27.680 I hope you're in the course of the next 15, 20 minutes. 00:27.840 --> 00:32.800 I'm really talking about this little random block of code that attends to turn normal English 00:32.800 --> 00:35.360 into grammatically correct English. 00:35.360 --> 00:39.920 And I know the joke is that they know there's British English and there's wrong. 00:39.920 --> 00:43.560 That's not directed towards the Americans, that's just fact. 00:43.560 --> 00:47.560 So I'm going to be talking about the origins of this little project. 00:47.560 --> 00:49.920 It's all up and get hub slides at the end. 00:49.920 --> 00:57.280 The problem I was trying to solve and the solutions that I adopted to solve that problem. 00:57.280 --> 01:01.920 It will be focused on some of the earlier stuff, but I should probably do who am I? 01:01.920 --> 01:04.840 What have I done to deserve a place on this stage? 01:04.840 --> 01:08.160 Or is the slides should be called the ego slide? 01:08.160 --> 01:10.480 This is where the speaker blides about themselves for 10 minutes. 01:10.480 --> 01:14.200 Well, everyone else thinks, can I get to the other deaf room I wanted to go to? 01:14.200 --> 01:15.880 Well, basically, I'm a geek. 01:15.880 --> 01:22.080 I've written some books, I write some music and someone recognized me as a magician earlier. 01:22.080 --> 01:25.240 I was doing it on the main stage a couple of years ago in the magic show. 01:25.240 --> 01:28.240 Does anyone see me do stuff at first and then before? 01:28.240 --> 01:30.240 And you came back? 01:30.240 --> 01:32.240 You're blabbering, I thought. 01:32.240 --> 01:38.240 So I've done a whole lot of crap, but the important thing is what's not on this slide. 01:38.240 --> 01:43.240 Yes, I've made games and but I'm not a translator, my English is awful. 01:43.240 --> 01:47.240 That's why I have editors to sort my books out for me. 01:47.240 --> 01:52.240 This is all about, well, if I can do this, anyone can. 01:52.240 --> 01:57.240 It should not be beyond the width of human beings to do just decent language. 01:57.240 --> 02:04.240 So this project started in about 2018, 2017 with that thing. 02:04.240 --> 02:10.240 Alexa, I was at a JavaScript meeting because JavaScript is the one true language. 02:10.240 --> 02:14.240 And no one's picked me up on that, okay, that's a different pattern all. 02:14.240 --> 02:18.240 No one's picked me up on that, okay, that's a different pattern all. 02:18.240 --> 02:21.240 Okay, so I thought, okay, I'll get one, I'll mess around with it. 02:21.240 --> 02:24.240 I went to various meetups, I wrote a lot of Alexa skills. 02:24.240 --> 02:27.240 And I found this was really interesting. 02:27.240 --> 02:33.240 I could harness a very complex technology into some very, very simple JavaScript. 02:33.240 --> 02:35.240 I thought, I liked that. 02:35.240 --> 02:39.240 And voice-in-touching, I thought, I like that. 02:39.240 --> 02:42.240 And voice-in-touching, I thought, I like that. 02:42.240 --> 02:45.240 And voice-in-touching, I thought, I like that. 02:45.240 --> 02:48.240 And voice-in-touching, I thought, I like that. 02:48.240 --> 02:53.240 And voice-in-touching, I thought, I was very interesting topic. 02:53.240 --> 02:57.240 Most of the time we have a screen with text on it, we can refer back to the text, if we forget. 02:57.240 --> 03:00.240 With the voice-in-touching, you can't. 03:00.240 --> 03:02.240 It plays, the voice ends. 03:02.240 --> 03:05.240 And you've got to remember everything that was said. 03:05.240 --> 03:07.240 I thought that was quite intriguing. 03:07.240 --> 03:10.240 So I started playing around with this, I wrote things that would tell you what they, 03:10.240 --> 03:12.240 the bins are taken out. 03:12.240 --> 03:15.240 Do you, everyone can remember what day they've got to take the bins out? 03:15.240 --> 03:16.240 No. 03:16.240 --> 03:19.240 Well, I don't need to either, because the person next to me, 03:19.240 --> 03:21.240 they take their bins out the day before, and I can see them. 03:21.240 --> 03:25.240 They are quite literally a bin fluencer. 03:25.240 --> 03:28.240 OK, never mind. 03:28.240 --> 03:31.240 So I thought, OK, I can do this JavaScript thing, 03:31.240 --> 03:35.240 and I'll do it as a service, because everything is as a service. 03:35.240 --> 03:38.240 You know, it's set software, as a service. 03:38.240 --> 03:40.240 You know, oh, Cracky. 03:40.240 --> 03:41.240 Anyone actually like AI here? 03:41.240 --> 03:43.240 Am I allowed to harass about AI in this room? 03:43.240 --> 03:44.240 Yes. 03:44.240 --> 03:45.240 Good. 03:45.240 --> 03:46.240 OK. 03:46.240 --> 03:47.240 I like Cracky as a service. 03:47.240 --> 03:48.240 It's like, oh, OK. 03:48.240 --> 03:49.240 Everything's as a service. 03:49.240 --> 03:50.240 I thought, OK, great. 03:50.240 --> 03:52.240 Well, this is a service thing. 03:52.240 --> 03:53.240 You make an API call. 03:53.240 --> 03:54.240 It fixes your grammar. 03:54.240 --> 03:56.240 It spits out the correct grammar. 03:56.240 --> 04:00.240 And this would do it in both English and as a speech thing. 04:00.240 --> 04:03.240 And that was an intentional choice. 04:03.240 --> 04:06.240 Because speech is different to words. 04:06.240 --> 04:08.240 And there were a lot of subtle things in speech 04:08.240 --> 04:11.240 that we just don't generally acknowledge. 04:11.240 --> 04:13.240 And hopefully, I'm going to come through some of them. 04:13.240 --> 04:15.240 And in 2025, that's last year. 04:15.240 --> 04:19.240 I know it feels like a lifetime ago, but it's only 30 odd days. 04:19.240 --> 04:23.240 And I thought, well, I'd written this thing for Alexa. 04:23.240 --> 04:26.240 Amazon said, hey, we really like this idea of this grammar 04:26.240 --> 04:27.240 as a service thing. 04:27.240 --> 04:31.240 We would like to promote you on our blog. 04:31.240 --> 04:32.240 OK. 04:32.240 --> 04:33.240 Are you paying me money? 04:33.240 --> 04:34.240 No, no, no. 04:34.240 --> 04:36.240 But we'll put a mention on our blog. 04:36.240 --> 04:38.240 No, I'm open-sourcing it. 04:38.240 --> 04:39.240 There's no money involved. 04:39.240 --> 04:40.240 It's open-sourcing. 04:40.240 --> 04:42.240 So I open-sourcing a whole lot last year. 04:42.240 --> 04:44.240 It's been used in a number of things. 04:44.240 --> 04:47.240 I don't know whether it's useful to anyone else except me. 04:47.240 --> 04:51.240 Because I get annoyed by a lot of bad, groundbreaking things. 04:51.240 --> 04:55.240 And that's someone who's not very good at English themselves. 04:55.240 --> 04:58.240 You know, my second language is drunkenese. 04:58.240 --> 05:02.240 You know, things like that says you have one email open-backed S. 05:02.240 --> 05:04.240 Isn't that just lazy? 05:04.240 --> 05:06.240 That should not be, I think. 05:06.240 --> 05:08.240 You pick up a axe. 05:08.240 --> 05:10.240 It's an axe. 05:10.240 --> 05:11.240 Look at it. 05:11.240 --> 05:13.240 Are we looking at a vowel here? 05:13.240 --> 05:15.240 Is this a vowel sound? 05:15.240 --> 05:17.240 And again, I mention it comes from Alexa. 05:17.240 --> 05:20.240 Comes from voice interfaces. 05:20.240 --> 05:22.240 What's this word, by the way? 05:22.240 --> 05:23.240 What's this? 05:23.240 --> 05:26.240 Oh, now you know there's a trick, don't you? 05:26.240 --> 05:29.240 Yeah, that's not AI. 05:29.240 --> 05:30.240 That's, ah. 05:30.240 --> 05:33.240 That's a free-toed slice. 05:33.240 --> 05:36.240 It's a perfectly valid word in scrabble. 05:36.240 --> 05:40.240 You can get up to 35 points with that word, but it's not AI. 05:40.240 --> 05:44.240 And if you throw that into a speaks synthesizer, it gets it wrong every time. 05:44.240 --> 05:48.240 So you need somewhere you're actually fixing just normal plain text 05:48.240 --> 05:51.240 when it goes out to the other end. 05:51.240 --> 05:54.240 What do you think happens when you feed this thing in? 05:54.240 --> 05:56.240 It is 30 minutes past five. 05:56.240 --> 05:58.240 No, it's not 30. 05:58.240 --> 06:00.240 That's how we say we say 30. 06:00.240 --> 06:01.240 We don't say 30. 06:02.240 --> 06:05.240 So when it comes out of the speaks thing, that feels very wrong. 06:05.240 --> 06:08.240 Us does Angela Merkel. 06:08.240 --> 06:10.240 Doesn't it great when it says Angela? 06:10.240 --> 06:12.240 That is Angela, but it's not his anger. 06:12.240 --> 06:14.240 So you've got to fix it. 06:14.240 --> 06:16.240 I always like this one. 06:16.240 --> 06:19.240 It's Polish or Polish. 06:19.240 --> 06:23.240 Maybe as a writer, I've just got sloppy and I haven't capitalised it. 06:23.240 --> 06:27.240 When it's a correct form there, you've got to get it and, you know, 06:27.240 --> 06:29.240 what language is this? 06:29.240 --> 06:31.240 It's not C sharp. 06:31.240 --> 06:34.240 It's C hash. 06:34.240 --> 06:37.240 Just the natural way it gets possessor C hash. 06:37.240 --> 06:39.240 But that's okay. 06:39.240 --> 06:43.240 Everyone that programs in this language is, you know, they all wear glasses 06:43.240 --> 06:45.240 because they can't C sharp. 06:45.240 --> 06:50.240 You kick me off the stage already. 06:50.240 --> 06:53.240 I'm going to annoy you with my question. 06:53.240 --> 06:56.240 Is it not C octauthor? 06:57.240 --> 06:58.240 Okay. 06:58.240 --> 07:00.240 It's one of those, is it? 07:00.240 --> 07:08.240 So there's a lot of cases where just the text doesn't quite convey all the information 07:08.240 --> 07:10.240 that it should convey. 07:10.240 --> 07:15.240 So I thought, well, either we can teach everyone to write AI this way. 07:15.240 --> 07:18.240 Or we could kind of kind of fix things. 07:18.240 --> 07:22.240 Because there is a book called The Chicago Manual of Style. 07:22.240 --> 07:25.240 And obviously when I say style, I don't mean dress style. 07:25.240 --> 07:28.240 Clearly, I've got read a book on style. 07:28.240 --> 07:29.240 But the shirt is good. 07:29.240 --> 07:30.240 You like the shirt? 07:30.240 --> 07:31.240 Yes. 07:31.240 --> 07:36.240 If my wife is watching, see, they like the shirt. 07:36.240 --> 07:42.240 Style in this sense means, how do you write proper like? 07:42.240 --> 07:45.240 You know, if you have a number, if it's below 10, 07:45.240 --> 07:48.240 you will actually write out the word for the number. 07:48.240 --> 07:50.240 If it's above 10, you write it out in digits. 07:50.240 --> 07:54.240 These very simple things is apparently quite hard. 07:55.240 --> 07:58.240 And I'm kind of a moderately native speaker of English. 07:58.240 --> 08:01.240 But it is still very difficult to make sure you get all the things right. 08:01.240 --> 08:03.240 So let's pass that off to a computer. 08:03.240 --> 08:07.240 And by a computer, I do not mean AI. 08:07.240 --> 08:10.240 What is the point of spending 14 billion a year on AI 08:10.240 --> 08:14.240 when you can just write 12 lines of JavaScript to do the same thing? 08:14.240 --> 08:18.240 So the solutions are basically, don't create a problem in the first place. 08:18.240 --> 08:20.240 This means everyone that writes code, 08:20.240 --> 08:25.240 this to be fluent in both language and in English. 08:25.240 --> 08:28.240 That's going to be expensive, that's going to be time difficult. 08:28.240 --> 08:30.240 So that's going to be a tricky thing. 08:30.240 --> 08:31.240 We can fix it manually. 08:31.240 --> 08:34.240 We can have a bunch of people just going through all the code, 08:34.240 --> 08:36.240 checking for all the grammar. 08:36.240 --> 08:38.240 Again, expense is not going to happen. 08:38.240 --> 08:40.240 We can ignore the problem. 08:40.240 --> 08:42.240 Most people ignore the problem. 08:42.240 --> 08:45.240 Every now and again, I'll see another piece of software. 08:45.240 --> 08:47.240 I'm going, no, another one. 08:47.240 --> 08:49.240 They're just ignored it. 08:49.240 --> 08:53.240 My attempt is, can we kind of find some way to automate this 08:53.240 --> 08:57.240 so that it doesn't just grind your gears? 08:57.240 --> 09:03.240 So, these are the types of things which the library automate. 09:03.240 --> 09:05.240 So it will look for numbers. 09:05.240 --> 09:08.240 And if there are numbers which is a plural number, 09:08.240 --> 09:10.240 it will make sure there's an S on the end. 09:10.240 --> 09:12.240 And this has been an employee melody for English, 09:12.240 --> 09:15.240 but it's been structured in such a way that any other language 09:15.240 --> 09:17.240 can just be plugged in. 09:17.240 --> 09:20.240 You write a little function that says, is this number a plural? 09:20.240 --> 09:21.240 Yes, no. 09:21.240 --> 09:22.240 Boolean, true false. 09:22.240 --> 09:24.240 At the end of the day, it will then fix it. 09:24.240 --> 09:25.240 Factions. 09:25.240 --> 09:28.240 If you write one slash two, 09:28.240 --> 09:30.240 the screen will say one slash two. 09:30.240 --> 09:31.240 It won't say a half. 09:31.240 --> 09:34.240 But we don't always say one half. 09:34.240 --> 09:36.240 So it will fix that sort of things. 09:36.240 --> 09:39.240 Symbols indefinite articles, the A's and the Anz, 09:39.240 --> 09:41.240 and certainly pronunciation, 09:41.240 --> 09:44.240 is certainly cursed into this as well, 09:44.240 --> 09:46.240 when it could be ambiguous. 09:46.240 --> 09:48.240 So what did he do? 09:48.240 --> 09:51.240 Well, it's JavaScript. 09:51.240 --> 09:55.240 You bring in the library, you say fix it. 09:55.240 --> 09:59.240 Throwing the text, make reference that there's going to be a number there, 09:59.240 --> 10:03.240 and it will just do the work. 10:03.240 --> 10:07.240 These are the different options you don't have to put every option in there. 10:07.240 --> 10:09.240 I want the style formatting. 10:09.240 --> 10:12.240 I want the plowels, but don't do the word number thing. 10:12.240 --> 10:14.240 The number is less than 10, you write it out. 10:14.240 --> 10:17.240 Sometimes that's not always what you want, 10:17.240 --> 10:19.240 so everything is switch on it lovable. 10:19.240 --> 10:21.240 Which can be fun. 10:21.240 --> 10:24.240 And then it looks like this, yes, it wraps around a bit. 10:24.240 --> 10:27.240 But all the code does is, things in a sentence, 10:27.240 --> 10:30.240 splits it up, looks at everything in there, and goes, 10:30.240 --> 10:34.240 okay, I'm going to process every single word in this sentence. 10:34.240 --> 10:37.240 I'm going to fix it up. 10:37.240 --> 10:41.240 And it's done in this way, so that you can add additional processes. 10:41.240 --> 10:46.240 So if your language is not English, and it has things which are not one of these five, 10:46.240 --> 10:49.240 it just sorts it out for you. 10:49.240 --> 10:51.240 So I'm using a number of other people's libraries, 10:51.240 --> 10:54.240 because I'm far too lazy to write my own. 10:54.240 --> 10:57.240 The pluralized library is really handy. 10:57.240 --> 10:59.240 You know the plowel of sheep. 10:59.240 --> 11:01.240 It's sheep. 11:01.240 --> 11:04.240 The plowel of person is people. 11:04.240 --> 11:07.240 This is not sticking us on the end. 11:07.240 --> 11:09.240 English has been such an annoying language, 11:09.240 --> 11:12.240 but it's everything has got an exception somewhere. 11:12.240 --> 11:14.240 And I'm sure I'm not the only one that gets annoyed by that. 11:14.240 --> 11:17.240 So there are other libraries, luckily out there that do this stuff. 11:17.240 --> 11:21.240 There's libraries that convert digits into actual numbers, 11:21.240 --> 11:25.240 spoken words, numbers into words, words that actually have silent 11:25.240 --> 11:29.240 h's that then change the ways, an hour or eight on this note, 11:29.240 --> 11:31.240 and on this person. 11:31.240 --> 11:35.240 Ooh, which I think I'm roughly almost getting now. 11:35.240 --> 11:38.240 And all the initialisms. 11:38.240 --> 11:42.240 All of this stuff, a normal speech synthesizer thing, 11:42.240 --> 11:44.240 is just going to look at that and go, 11:44.240 --> 11:46.240 A, C, or A, or A, 11:46.240 --> 11:49.240 is probably not going to do what you mean it to. 11:49.240 --> 11:52.240 And the fun one, all the capitalisms. 11:52.240 --> 11:54.240 These words change their meaning, 11:54.240 --> 11:57.240 whether this is a capital or not. 11:57.240 --> 12:00.240 So in lower case, it's August. 12:00.240 --> 12:02.240 A capital A is August. 12:02.240 --> 12:04.240 Completely different word. 12:04.240 --> 12:06.240 Polish we did earlier, Polish and Polish, 12:06.240 --> 12:08.240 and Job and Job again. 12:08.240 --> 12:11.240 I had a copy of these from Wikipedia. 12:11.240 --> 12:12.240 I'm not that smart. 12:12.240 --> 12:16.240 These are the sort of things that you just have to worry about, 12:16.240 --> 12:18.240 particularly with voice interfaces. 12:18.240 --> 12:20.240 And if you're thinking, well, I'm never going to do a voice 12:20.240 --> 12:21.240 interface thing ever in my life. 12:21.240 --> 12:23.240 So like, well, why not? 12:23.240 --> 12:27.240 At some point, we have people who just can't lead the screen. 12:27.240 --> 12:30.240 I am now getting to that age where I can't see the screen as well 12:30.240 --> 12:32.240 as I used to when I was a younger person. 12:32.240 --> 12:33.240 I was going to say a younger person. 12:33.240 --> 12:35.240 I think I've ever been young in my life. 12:35.240 --> 12:36.240 I was born old. 12:36.240 --> 12:38.240 But I'm getting to the point. 12:38.240 --> 12:40.240 It's like, I need these helping. 12:40.240 --> 12:43.240 I need the device to say things to me if I can't lead the screen. 12:43.240 --> 12:46.240 So at some point, screen readers are going to become much more prevalent, 12:46.240 --> 12:48.240 especially as our generation of people get older. 12:48.240 --> 12:49.240 And we suddenly realize, 12:49.240 --> 12:52.240 damn, I actually do need screen readers now. 12:52.240 --> 12:56.240 So these sort of things are going to become my think more permanent. 12:56.240 --> 12:59.240 So as an example, what is that? 12:59.240 --> 13:02.240 That's a fup agent. 13:03.240 --> 13:07.240 Because, I mean, how many, I'm not going to test anybody. 13:07.240 --> 13:08.240 There's no quiz at the end of this. 13:08.240 --> 13:10.240 The quiz was last night. 13:10.240 --> 13:12.240 Was anyone there for the quiz last night? 13:12.240 --> 13:13.240 No? 13:13.240 --> 13:14.240 Well, you should have been there. 13:14.240 --> 13:15.240 It was brilliant. 13:15.240 --> 13:17.240 We gave out free beer to everyone. 13:17.240 --> 13:18.240 You missed it. 13:18.240 --> 13:19.240 Sorry. 13:19.240 --> 13:21.240 It was me again, yeah. 13:21.240 --> 13:25.240 So there were about four problems with that sentence. 13:25.240 --> 13:28.240 And if you flow it into the library, it fixes it. 13:28.240 --> 13:30.240 That's the correct output. 13:30.240 --> 13:32.240 It's not a, it's an, because it's f. 13:32.240 --> 13:34.240 And that's a vowel sound. 13:34.240 --> 13:36.240 It's a capital A. 13:36.240 --> 13:38.240 And it has to have a full stop at the end. 13:38.240 --> 13:40.240 And FPI agent. 13:40.240 --> 13:43.240 And this is the SSML, which is the speech version of it. 13:43.240 --> 13:47.240 So it will actually spell out FPI. 13:47.240 --> 13:49.240 This stuff is really quite handy. 13:49.240 --> 13:51.240 And look at teach me. 13:51.240 --> 13:53.240 This is either the most useful thing of the library, 13:53.240 --> 13:55.240 or the most pretentious. 13:55.240 --> 13:59.240 I put a bit in the library that man explains to you where you went wrong. 13:59.240 --> 14:04.240 FPI is an initialism, not an acronym. 14:04.240 --> 14:05.240 A initialism. 14:05.240 --> 14:06.240 There's a difference. 14:06.240 --> 14:08.240 And should be written with dots like this. 14:08.240 --> 14:11.240 And it uses the example in the output as well. 14:11.240 --> 14:16.240 So it kind of handles itself as a teaching tool for everyone else. 14:16.240 --> 14:20.240 And it gives it in text, standard plane, HTML along with the SSML. 14:20.240 --> 14:22.240 So that's all good. 14:22.240 --> 14:24.240 One fairly nice example. 14:24.240 --> 14:27.240 The next steps on this, I truly don't know. 14:27.240 --> 14:30.240 I wrote this stuff because I liked it. 14:30.240 --> 14:33.240 I needed it when I was doing Alexa stuff. 14:33.240 --> 14:35.240 I needed the library. 14:35.240 --> 14:38.240 Well, now I don't do Alexa stuff anymore. 14:38.240 --> 14:40.240 Do I still need the library? 14:40.240 --> 14:41.240 I don't know. 14:41.240 --> 14:42.240 I think it's an interesting problem. 14:42.240 --> 14:44.240 I think it's a worthwhile problem to look at. 14:44.240 --> 14:47.240 So I'm here to sort of say, look, is this fun? 14:47.240 --> 14:48.240 Is this interesting? 14:48.240 --> 14:49.240 Is this even useful? 14:49.240 --> 14:52.240 If not, yeah, we have a nice time. 14:52.240 --> 14:56.240 So if it's useful, it's great if not. 14:56.240 --> 14:57.240 Fine. 14:57.240 --> 15:00.240 There's a number to do in the code. 15:00.240 --> 15:03.240 When I wrote this, because it wasn't written as open source, 15:03.240 --> 15:06.240 some of the code is a bit crap. 15:06.240 --> 15:12.240 When I open source this, I suddenly realize that I felt the word sentence incorrectly. 15:12.240 --> 15:17.240 You know how bad it is when you're writing a piece of software that does language and grammar correction 15:17.240 --> 15:21.240 and you've misspelled the most used word in your product. 15:22.240 --> 15:24.240 So I had to fix a spelling of sentence. 15:24.240 --> 15:25.240 And a few other things. 15:25.240 --> 15:26.240 There's a few other bits there. 15:26.240 --> 15:29.240 It's only JavaScript because it was written as a service. 15:29.240 --> 15:33.240 But there's no reason why you can't just copy and paste this stuff and just change a few bits 15:33.240 --> 15:34.240 to make it work elsewhere. 15:34.240 --> 15:37.240 And of course, there's Job's job. 15:37.240 --> 15:39.240 Is anyone aware of this? 15:39.240 --> 15:44.240 Oh, this is either going to really entertain you or bore you senseless. 15:44.240 --> 15:47.240 This is from Richard Lederer's word circus. 15:47.240 --> 15:50.240 I don't know if anyone's ever read this, probably not. 15:50.240 --> 15:55.240 But in August, an August Patriot was reading an ad in Reading, Massachusetts. 15:55.240 --> 16:01.240 Long-suffering Job secured a job to polish piles of Polish brass. 16:01.240 --> 16:06.240 You try getting a machine to correctly read that. 16:06.240 --> 16:12.240 I would like to be able to have libraries that could take that and go, oh yeah, I know what to do with that. 16:12.240 --> 16:14.240 And don't say AIs a solution. 16:14.240 --> 16:16.240 AIs never the solution. 16:16.240 --> 16:20.240 I think nice confined pieces of code that just do one job. 16:20.240 --> 16:23.240 The proper Unix way might actually help. 16:23.240 --> 16:26.240 So I would like to be able to get that read perfectly. 16:26.240 --> 16:28.240 And the code almost gets it right. 16:28.240 --> 16:32.240 Without doing any special work, it gets most of that right. 16:32.240 --> 16:36.240 Because it understands the initialism, Polish and Job and Job and Reading, 16:36.240 --> 16:38.240 versus Reading. 16:38.240 --> 16:41.240 That one, that's really good fun. 16:41.240 --> 16:45.240 Because I'm from England, there is a town called Reading, not far from me. 16:45.240 --> 16:48.240 And I love just going there and just saying, well, you know, you people, 16:48.240 --> 16:53.240 there's a very sharp decline in the standard of reading. 16:53.240 --> 16:58.240 Okay, for the three people, it was worth it. 16:58.240 --> 17:03.240 So my conclusion is, groundwork is actually really quite hard. 17:03.240 --> 17:08.240 When you look at what's going on underneath the language, there are problems there. 17:08.240 --> 17:10.240 So it's not surprising. 17:10.240 --> 17:14.240 We just kind of forget about it and we push it and we just, oh, someone will fix it later. 17:14.240 --> 17:18.240 Let's see if we can fix it now. Let's get it right. 17:18.240 --> 17:20.240 It's definitely hard with dynamic content. 17:20.240 --> 17:23.240 Anytime there's dynamic content in there, go over. 17:23.240 --> 17:26.240 You've got to start thinking about this. 17:26.240 --> 17:29.240 And it's not just about the code, it's about the language underneath it. 17:29.240 --> 17:31.240 I try and think voice first. 17:31.240 --> 17:34.240 I like the idea the computer is talking to me. 17:34.240 --> 17:38.240 The voice is what's coming to me, not the text. 17:38.240 --> 17:43.240 So I think, okay, is there a way of understanding what's being said here on screen? 17:44.240 --> 17:46.240 And could we get this to other languages? 17:46.240 --> 17:48.240 You know, I am very fully British. 17:48.240 --> 17:51.240 I speak, which almost know what the language is. 17:51.240 --> 17:55.240 My French is bad, my German is bad, my Japanese is awful. 17:55.240 --> 17:59.240 So I am very much the stereotypical English guy. 17:59.240 --> 18:02.240 So I would love to see this in other languages. 18:02.240 --> 18:03.240 And there's no reason why not. 18:03.240 --> 18:06.240 It's built in such a way that it can be ported. 18:06.240 --> 18:09.240 So with that in mind, I can do any questions that anyone has. 18:09.240 --> 18:10.240 I will update my scorecard. 18:10.240 --> 18:11.240 There we go. 18:11.240 --> 18:12.240 30th talk. 18:12.240 --> 18:15.240 I've given this conference over 26 years. 18:15.240 --> 18:17.240 I'll buy and does it show. 18:17.240 --> 18:19.240 But otherwise, that's all for me. 18:19.240 --> 18:20.240 I can take questions. 18:20.240 --> 18:21.240 Thank you. 18:21.240 --> 18:22.240 Good night. 18:22.240 --> 18:31.240 I have a question, can we? 18:31.240 --> 18:39.240 Or a point that might be either very funny or not. 18:39.240 --> 18:44.240 When you show me the explanation, I think there is something you don't. 18:45.240 --> 18:46.240 Yeah, if I want? 18:46.240 --> 18:47.240 Yeah. 18:47.240 --> 18:54.240 On the end of teaching me, there is FBI. 18:54.240 --> 18:57.240 On the, or yes, there, okay. 18:57.240 --> 19:01.240 Okay, submit a PR. 19:01.240 --> 19:03.240 We'll fix it in the next version. 19:03.240 --> 19:04.240 Okay. 19:04.240 --> 19:05.240 It's supposed to jump in now. 19:05.240 --> 19:07.240 I'm sorry. 19:07.240 --> 19:08.240 Okay. 19:08.240 --> 19:11.240 So real question. 19:11.240 --> 19:12.240 Comment first. 19:12.240 --> 19:15.240 Then the question that that was a context leads to. 19:15.240 --> 19:20.240 I work with non-anglophone people writing. 19:20.240 --> 19:23.240 Anglophone comments in code. 19:23.240 --> 19:31.240 How extensive could this grammar as a service solution be made to correct languages made by non-native speakers? 19:31.240 --> 19:33.240 Can I see there are a series of corrections. 19:33.240 --> 19:37.240 But I don't know how many corrections you could make. 19:37.240 --> 19:38.240 Yeah. 19:38.240 --> 19:41.240 There's no reason that depending on your. 19:41.240 --> 19:44.240 So the question is about. 19:44.240 --> 19:48.240 See what I did there. 19:48.240 --> 19:53.240 So the question is about coders who are non-native speakers and how they could fix it while 19:53.240 --> 19:54.240 they're coding. 19:54.240 --> 19:56.240 And they answer his well, why not? 19:56.240 --> 19:59.240 There's no reason you couldn't have this service running either on your local machine. 19:59.240 --> 20:01.240 It's very, very lightweight. 20:01.240 --> 20:02.240 It's JavaScript. 20:02.240 --> 20:03.240 A couple hundred lines. 20:03.240 --> 20:07.240 You can just have that running and your editor can just say, okay, here's the code. 20:08.240 --> 20:09.240 What do you think of it? 20:09.240 --> 20:14.240 And then it can throw back a fixed grammar version for you while you're still there. 20:14.240 --> 20:16.240 It's something you'd have to set up. 20:16.240 --> 20:18.240 But there's no reason why it couldn't. 20:18.240 --> 20:21.240 You know, it was built to be an API. 20:21.240 --> 20:23.240 So yes, go for it. 20:23.240 --> 20:24.240 You had a question. 20:24.240 --> 20:25.240 I think you as well. 20:25.240 --> 20:26.240 And there's another one on here. 20:26.240 --> 20:27.240 A real one. 20:27.240 --> 20:28.240 Really? 20:28.240 --> 20:29.240 Yeah. 20:29.240 --> 20:30.240 Good. 20:30.240 --> 20:31.240 Well, there's this thing. 20:31.240 --> 20:34.240 We have in a weblight called Intermediate Language. 20:34.240 --> 20:42.240 It's usually used that you have a lot of, well, mostly all of the software is written in English. 20:42.240 --> 20:45.240 No matter that nobody is English native. 20:45.240 --> 20:47.240 And even English native. 20:47.240 --> 20:52.240 Well, even British people are messing the grammar as you know. 20:52.240 --> 20:53.240 Yeah. 20:53.240 --> 21:00.240 So mostly it's used as developer English, meaning that you have, before you let 21:00.240 --> 21:03.240 translate this all machines to do the translation. 21:03.240 --> 21:06.240 You use Intermediate, developer English. 21:06.240 --> 21:12.240 So you as writer of the code can do something that's very understandable. 21:12.240 --> 21:17.240 Then that's translated by reviewers to English. 21:17.240 --> 21:20.240 That's used as a source for all the translations. 21:20.240 --> 21:29.240 And we have, or Weblight has plenty of machine translation and machines integrated to do 21:29.240 --> 21:32.240 machine translation general. 21:32.240 --> 21:35.240 But probably this could be very specific. 21:35.240 --> 21:37.240 And we should talk to each other. 21:37.240 --> 21:44.240 Because do you think that you would like to modify this a little bit to be a provider for 21:44.240 --> 21:51.240 machine translation from English to good English form? 21:51.240 --> 21:52.240 Why not? 21:52.240 --> 21:55.240 So the question is, use this elsewhere. 21:55.240 --> 21:57.240 And the answer is, why not? 21:58.240 --> 22:01.240 I don't know what the solution is here. 22:01.240 --> 22:06.240 As I don't think English is the one and true language. 22:06.240 --> 22:10.240 But unfortunately it has become the effect of language of the planet. 22:10.240 --> 22:17.240 So if we can at least use decent English, wherever it's being used, then yeah. 22:17.240 --> 22:22.240 I'll try and I'll hack change and mess around with this as much as necessary. 22:22.240 --> 22:24.240 So get good stuff out there. 22:24.240 --> 22:26.240 So you know, let's face it. 22:26.240 --> 22:27.240 I'm an employable. 22:27.240 --> 22:29.240 So I've nothing else to do this week. 22:29.240 --> 22:33.240 So let's see, let's see what's wrong and move fix it. 22:33.240 --> 22:35.240 So we have to get in one scene. 22:35.240 --> 22:38.240 There are those are brought in a sentence. 22:38.240 --> 22:42.240 Because you have to see, I have to you have to get in one scene. 22:42.240 --> 22:46.240 But a gentleman of Dutch people translating laterally is, 22:46.240 --> 22:49.240 you have one scene for get in. 22:49.240 --> 22:52.240 So I've gone vastly in Chinese. 22:52.240 --> 22:55.240 It's more, more completely conversely, 22:55.240 --> 22:59.240 inversely, on the wall in the same place. 22:59.240 --> 23:00.240 So it is. 23:00.240 --> 23:03.240 Yeah, Germany's time-manor places in order. 23:03.240 --> 23:05.240 You've got to get that right. 23:05.240 --> 23:08.240 Yeah, I remember that for my coordinates. 23:08.240 --> 23:10.240 You had a question still. 23:10.240 --> 23:11.240 Yeah. 23:11.240 --> 23:14.240 You were talking about screenwriters. 23:14.240 --> 23:15.240 Yeah. 23:15.240 --> 23:20.240 You can't necessarily go back and read something because it's voice. 23:20.240 --> 23:24.240 So there's a, I don't really know the tech dialects or screenwriters, 23:24.240 --> 23:31.240 but it doesn't serve us to break up one of the sentences. 23:31.240 --> 23:35.240 So does the library break up the sentence? 23:35.240 --> 23:38.240 Not really, it takes the whole thing in. 23:38.240 --> 23:39.240 It looks at everything. 23:39.240 --> 23:43.240 Try to apply some context to that sentence and go, 23:43.240 --> 23:46.240 okay, I'm now changing the whole thing. 23:46.240 --> 23:53.240 So it wouldn't currently be a lot of use in a live translation environment, 23:53.240 --> 23:55.240 but it could be used if you said, well, 23:55.240 --> 23:59.240 let's just look at the next comma, pass that to the library, 23:59.240 --> 24:02.240 get that to fix it and then do the next bit. 24:02.240 --> 24:05.240 So it is hackable in that way and it's, 24:05.240 --> 24:07.240 and when you've got that sort of library, 24:07.240 --> 24:08.240 it's up to you what you feed it. 24:08.240 --> 24:12.240 So you could just say, well, I'll give you the 10 words I've got, 24:12.240 --> 24:16.240 and then maybe it will make something good after it. 24:16.240 --> 24:18.240 Yeah, something I actually have, I forgot, 24:18.240 --> 24:21.240 I was giving a talk about Eliza, the AI chatbot Eliza, 24:21.240 --> 24:22.240 just prior to this. 24:22.240 --> 24:24.240 And I forgot to mention that Eliza, 24:24.240 --> 24:26.240 if you have one heard of Eliza, 24:26.240 --> 24:28.240 the AI chatbot from 60 years ago, 24:28.240 --> 24:31.240 yeah, well, that doesn't process the whole sentence. 24:31.240 --> 24:33.240 It stops at the first comma, 24:33.240 --> 24:35.240 all the first word, but. 24:36.240 --> 24:38.240 And it just does the bit before, 24:38.240 --> 24:40.240 and ignores the rest of it. 24:40.240 --> 24:42.240 It's like when you send an email to your boss 24:42.240 --> 24:44.240 and they only ask the first question on your list, 24:44.240 --> 24:46.240 that's what it does. 24:46.240 --> 24:48.240 So you could have dropped that mechanism 24:48.240 --> 24:50.240 and just say, well, I'm just going to look for a comma, 24:50.240 --> 24:52.240 send that off, get that fixed. 24:52.240 --> 24:54.240 That career solution. 24:54.240 --> 24:56.240 I saw another hand wave, yes. 24:56.240 --> 24:58.240 You were talking about, like, 24:58.240 --> 25:00.240 attracting, which, like, written, which is correct, 25:00.240 --> 25:02.240 but I want to understand how you would have 25:02.240 --> 25:04.240 to be Lord words, which English uses, 25:04.240 --> 25:06.240 like, English is not, like, 25:06.240 --> 25:09.240 consolidated if you knew you came and now. 25:09.240 --> 25:11.240 And there are a lot of Lord words, like, 25:11.240 --> 25:14.240 we understand, like, Hasbro, 25:14.240 --> 25:16.240 but who I know that, which are using English 25:16.240 --> 25:18.240 and which are kind of used in legal terms, 25:18.240 --> 25:21.240 but these days, how would you use it? 25:21.240 --> 25:25.240 So the library does have some loan words in it. 25:25.240 --> 25:28.240 So the idea of how do we handle loan words, 25:28.240 --> 25:31.240 they are quite literally hard-coded in there. 25:31.240 --> 25:33.240 And I'm saying, these are the ones you need to pay 25:33.240 --> 25:36.240 special attention to, but it's not complete. 25:36.240 --> 25:39.240 So hopefully over time that will grow, 25:39.240 --> 25:42.240 maybe it's the case of splitting out the library. 25:42.240 --> 25:44.240 So we have one library that says, 25:44.240 --> 25:46.240 okay, I'm going to do all the numbers 25:46.240 --> 25:47.240 to words calculation. 25:47.240 --> 25:49.240 I'm going to do all the pluralization, 25:49.240 --> 25:51.240 this library will do loan words. 25:51.240 --> 25:52.240 Maybe that's the way it goes. 25:52.240 --> 25:54.240 Maybe this project just has to go away, 25:54.240 --> 25:58.240 so that the other libraries take care of the big bulk of loan words. 25:58.240 --> 26:03.240 Because as you say, English is quite literally all loan words. 26:03.240 --> 26:05.240 I don't know, I don't know how many words in English 26:05.240 --> 26:07.240 are actually English words anymore. 26:07.240 --> 26:10.240 You know, not to misquote George W. Bush, 26:10.240 --> 26:12.240 but the fans have no word for entrepreneur. 26:12.240 --> 26:14.240 That kind of thing. 26:14.240 --> 26:17.240 We have literally stolen everything. 26:17.240 --> 26:19.240 That's why the British Museum has got some really good 26:19.240 --> 26:21.240 Elgin marbles in it. 26:22.240 --> 26:25.240 Okay, I'm not being allowed back in the British library. 26:25.240 --> 26:26.240 Am I? 26:26.240 --> 26:28.240 Okay, any other questions? 26:28.240 --> 26:30.240 I know I've got about two minutes left to come 26:30.240 --> 26:32.240 to my timekeeper at the front. 26:32.240 --> 26:33.240 No? 26:33.240 --> 26:35.240 All in which case? 26:35.240 --> 26:36.240 Thank you all again. 26:36.240 --> 26:37.240 Good night. 26:37.240 --> 26:39.240 Thank you.