One advantage of regular letters compared to this is that it's pretty easy to tell that even a fairly mangled A is still an A; a word written in dotsies loses a lot of its legibility as soon as it's not displayed in fairly large pixels across a very clear background. Viewing things at an angle? While it moves by? On a dirty or wet piece of paper? Forget about it.
It's a super fun idea, though. I would 100% put this in the background, or even the foreground, of some future scifi scape.
Me too, I didn't even mean to read your comment, but as soon as my eyes saw it: boom, processed and read. like a computer without virus scan or security or anything, to see the words was to run their executable on my wetware.
I actually remember to have had severe headaches when I was something like 4 or 5 and I experienced the passage from deliberate reading and automatic reading.
It was much worse being around in a car and just looking outside and unwillingly read all the signs because it added also elements of motion sickness to it...
Nope: automatic readers rely heavily on redundancy. QR codes are like 20% data, 80% error correction. OCR is still a difficult problem, despite the redundancy built in to the alphabet. There's no way a machine reader could tell apart characters that differ by one pixel
Yeah, it's a telling point that when we're asked to read passages of Dotsies, the font's sized up least three times (36px to 42px)—and so at least nine times the screen area taken—compared to the base 13px size used in the "How much better is it?" example.
You're going to need damn good eyes to read Dotsies at anything like 13px, and I suspect even the sharp-eyed are going to misread a lot. Size Dotsies text up to the point that people of average vision can read it as reliably as a normal font at a reasonable size, and I don't think you're going to get much, if any, greater information density.
I'm reminded of Speedtalk, a language from Robert Heinlein's story "Gulf", which had every combination of phoneme and delivery (pitch, length, etc.) represent a word in the Basic English vocabulary. So, every spoken "word" in the language is actually a full sentence. But, it takes intense futuristic training just to be able to speak vaguely intelligibly in it, because all the normal imprecisions of speaking (much less speaking in a new language) take any tiny error from "You used 'record' as a noun, but with a long 'e' and the stress on the second syllable." to "You have a hovercraft, and it's full of what???"
Now imagine trying to follow speech in such a language if the speaker's out of breath, alarmed, surrounded by background noise, etc. Just not enough redundancy for reliable communication.
Not to mention recognizing individual letters. The difference between “a” and “b” in Latin alphabet is drastic. In dotsies you could only tell them apart if other letters near them give you relative height of the dots to some baseline.
Dotsies seem to make a tradeoff where they optimize for higher information density at the cost of much higher error rate (or requirement for precision and accuracy if you will). I also definitely see your point about font sizes. Basically it’s a valiant attempt at a reimagined alphabet but I don’t think it actually achieves higher info density in the real world.
I like the idea of users being able to choose the font they prefer to read in. It's not very hard to do: you could configure your browser and your code editor and maybe Kindle and be able to read in the new font 90% of the time.
But Dotsies seems to be solving the wrong problem: using the fewest pixels. I think the right problem is being able to skim large amounts of text quickly, for which typical fonts are pretty good. You can find whether there are any numbers on a page, for instance, in well under a second.
I use it (well, OpenDyslexie) as my only font on desktop Firefox and it's been ok for me for a year or thereabouts. I did use it in iOS Pocket as well but they dropped it a while back.
Sighted braille teachers can read braille visually, even upside down, so they can read what's coming out of the student's brailler in real time. I had at least one teacher who could do that.
But of course, that's a specialized skill. To judge for yourself how conducive braille is to visual reading, when presented as black on white (or the opposite) instead of raised dots on a page, you can grab a braille font here:
Yes but that isn't meant for screen space saving or legibility - dotsies is optimizing for both.
They could maybe do better on legibility, or maybe it isn't that big a problem...the more I think on it, the more I get where the author was going (and feel they may have successfully reached it, or at least made strong progress)
this is significantly more compact than braille. braille is less compact than latin characters, and the point of this exercise was more compact information.
Braille has many of the same properties, and it could be worthwhile to train yourself to read it visually. If you went down that path, though, I think you'd get to a point where you'd say: hmm, if I'm going to start over I might as well make the letters one column of dots, so the words are roughly square-ish and can form more visually distinct shapes. The shape of Braille words wouldn't be much of an improvement over Latin - unnecessarily long and skinny. Which is probably a good fit for a finger but less optimal for an eye. And then there's that unnecessary sixth dot.
As to the originality, most computer science students have looked at a binary versions of ascii charts at one time or another. Turning them sideways and making a font out of it isn't discovering relativity. On the other hand, even Braille had its percursors (see your Wikipedia link), and Braille users don't reject it on that account, do they?
> Silicon Valley
You're assuming I'm from Silicon Valley. I am here now, but was actually in Ohio when I came up with Dotsies :)
I love the way it teaches you, but the other comments are correct.
It doesn't really make sense to optimize for screen space like this.
However, the teaching style along with a symbol alphabet that IS trying to solve for better legibility (without necessarily any allegiance to dyslexic fonts), could be pretty interesting.
In middle school when bored once I made my own symbolic alphabet, I found it really is extremely easy to learn a new symbol set for language you already know, so, changing away from the traditional symbols isn't even that massive of an endeavor.
I think they did do an OK job with using screen space well, but at the cost of legibility to a degree. That said, the more I think on it, maybe the author did too. Sure, individual letters aren't as legible, but the point is almost more that words become single characters, more like Mandarin. If those characters are legible, than it doesn't matter that the individual letters are not. If you don't know a word, you're already going to be individually inspecting the letters. (of course, it will still be mildly more difficult with dotsies)
This was my favourite aspect of dotsies: that words began to have familiar shapes. I stopped needing to parse certain words because their shape was so unique and memorable.
At the same time it would be interesting to see it implemented based on letter frequency rather than order in the alphabet.
Isn't that true for "normal" text too? Don't we skim text usually pretty quickly based on how the words look like, instead of reading every letter one by one?
> It doesn't really make sense to optimize for screen space like this.
However, the teaching style along with a symbol alphabet that IS trying to solve for better legibility ...
Optimizing for legibility as opposed to screen space is a distinction that isn't very meaningful in practice. Because if words aren't legible enough, people increase the font size until they are. If something is extremely legible (like latin letters at a 40pt font size) people will naturally tend to decrease the size. So screen space and legibility are largely different labels for the same concept. With some wiggle room but not that much.
just use Chinese, it’s optimized for screen space (each character occupies a square, and has complex meaning of a word, and people read their shapes, rather than stroke by stroke )
I think a key point of the language is that it points toward machine-reproduced printing and display only. Writing can still be Chinese, English or any other language. But for in-place printed labeling or digital display, something like this might be interesting.
It occurs to me that the suggestion is actually interesting from an information theoretic standpoint. I understand that Chinese as a language is relatively "telegraphic" to start with. Is there a quantifiable difference with English writing? How much is the language and how much the writing system?
Because the hubris of tech people to think something like "hey, trillions of humans have collectively decided over a few thousand years that the latin alphabet hits a pretty good sweet spot in human communication. I bet they didn't consider how readable it is, though. I can do better than all of them," is wearing thin.
This came up in the Robinhood checking thread a few days ago. "Do you really think that you're going to outsmart a financial institution that employs people like you (but smarter) because you happen to be better at math than your facebook friends who studied polysci?" The hubris of people in tech who think they're the only people who notice things is getting really tiresome.
Having said that, I think that's a slippery slope of a stance to take, because it stifles creativity and novel solutions, which history shows often come up when they're not looked for. But I think that's kind of the sentiment of embarrassment they were expressing: "groan, another techie is going to disrupt something that doesn't need disrupting."
Edit: To the people responding to this (and downvoting): this person asked for an explanation for why someone would feel embarrassment at this. I was trying to give what I thought was a reasonable explanation. I don't actually feel embarrassed by this, I was just trying to understand and articulate someone else's feelings on this. I'm sorry if the explanation didn't do it for you.
Robinhood is a weird example to bring up here. All the HNers who thought they knew better than a $6 billion financial company with an army of professionals turned out to be right, Robinhood got smacked down hard, and that top-rated comment in the initial thread now looks hilariously silly.
I completely fail to see the problem here. Let people try to disrupt whatever they want, what's the harm? It's extremely unlikely that dotsies will ever be anything other than a minor curiosity, but where's the problem? There's a one in a billion chance it could be the next evolution of the alphabet, and, anyway, what's bad about trying?
I don't see a reason for the negativity, because I don't see a single disadvantage in trying. If the author wants to waste their time, let them.
The difference from what was decided historically is that in the past, books had to be printed in a font chosen by the printer, so they chose one that the most people could read. Today, individuals can choose their own fonts to read in, so the legacy compatibility problem is greatly reduced.
Given how different some international alphabets are (Russian, Hebrew, Katakana) it's not self-evident that the conventional Latin glyphs are optimal.
> hey, trillions of humans have collectively decided over a few thousand years that the latin alphabet hits a pretty good sweet spot in human communication. I bet they didn't consider how readable it is, though. I can do better than all of them
I don't know all the fancy names for logical fallacies but this seems to be deliberately misrepresenting the argument. The website clearly states that the problem with the Latin alphabet is that it has to be able to be written efficiently as well as read. Your trillions of humans over a few thousand years have had to both read and write this alphabet. In the past few hundred years we've had technology that allows someone who made a huge investment to produce massive amounts of text without writing, and in the past few decades the ability to produce text without writing has become completely ubiquitous. Does it still make sense for this alphabet to optimize for hand writing? Are you arguing that the constraints that made the Latin (or any handwritten) alphabet have remained unchanged over thousands of years? They want to redesign an alphabet to fit modern constraints. We consume a lot of text on devices that need to have screens that fit in our pockets. Perhaps a two-thousand year old alphabet isn't still the optimal solution?
I also don't think the authors necessarily expect to disrupt anything. They just made something cool and put it on the internet. I'm happy people like this exist.
Yeah, text on computers and phones is so much different from handwritten or even printed text. We don't have to handwrite text that often anymore, so we don't have the constraints of how accurate our hands can move. Our screens have really high resolution and color (although using a script with colors would be problematic for color blind people), so we don't have the constraints of the paper medium. There's also the simple fact that Latin isn't the only writing system in existence, so how are we to know that Latin is the best writing system out of all the writing systems already in existence? Why not hànzì? Or hangul? Or Devanagari? Or the Arabic Script? Or even Canadian Aboriginal Syllabics? There are so many different scripts that it seems weird to me to focus on the Latin Script as the "pretty good sweet spot". That seems like the much bigger hubris to me. It's not like the Latin Script was scientifically created and proven to be efficient. It just organically evolved from what came before it, whether those changes were beneficial or not. And of course again, they didn't have screens back then.
> Because the hubris of tech people to think something like "hey, trillions of humans have collectively decided over a few thousand years that the latin alphabet hits a pretty good sweet spot in human communication. I bet they didn't consider how readable it is, though. I can do better than all of them," is wearing thin.
This isn't a unique aspiration of tech. Folks have been trying to achieve better othographies than the one we grafted on as a colonizer's comprimise for a long time. Projects like Shavian [0][1], Read Script (Quickscript) [2], and Deseret [3] come to mind as prior art in this space that predate the invention of the modern tech community and its current perceived class distinction.
I'd argue it's the supreme arrogance to assume that in a modern world populated with people who's average intelligence, nutrition and education was only accessible to the richest and most powerful 0.001% 200 years ago that NO progress can be made on a common linguistic platform. The classical alphabet we use has ALSO been revised and has literally hundreds of variants.
On a personal note, I don't at all regret the decision to learn to read and write in Shavian. Having a semi-private notation with which to take notes, form signatures, etc. The only thing I am sad about is that so few people in this community (and others) decided to put in the modest effort required to do the same. If I do occasionally offer thoughts in ·𐑖𐑱𐑝𐑾𐑯, 𐑐𐑰𐑐𐑤 𐑸 𐑩𐑯𐑜𐑮𐑰, 𐑮𐑧𐑕𐑧𐑯𐑑𐑓𐑳𐑤 𐑰𐑝𐑧𐑯! 𐑤𐑲𐑒 𐑞 𐑧𐑒𐑕𐑦𐑕𐑑𐑧𐑯𐑕 𐑝 𐑩𐑤𐑑𐑺𐑯𐑩𐑑𐑦𐑝 𐑹𐑔𐑪𐑜𐑮𐑩𐑓𐑰 𐑦𐑟 𐑩𐑯 𐑦𐑯𐑕𐑳𐑤𐑑. 𐑲 𐑛𐑪𐑯𐑑 𐑜𐑧𐑑 𐑦𐑑.
I'm debating if dotsies would be valuable to learn. My biggest complaint about it is that it's a reproduction of our current alphabet, and that means you'll end up with a lot of our weird di-and-trigraphs. If you're going to to the trouble of a con-script, why not use a phonetic base?
You seriously, seriously need to reconsider a significant part of your worldview if you can easily think of the reasons one ought to be "proud" of such an effort, but you can't think of the reasons one might be "embarrassed".
Without knowing it, you've taken a weirdly extreme and limited perspective - and deprived yourself of anything remotely close to a complete understanding.
I assume you're probably proud of your "optimism" or something here, but it's absolutely silly.
I feel like HN should implement a "positive-only" or "constructive-only" flag for posts, whether set by posters or moderators. Perhaps all Show HN. Start to experiment, see what sticks.
It is a pity an intellectually curious person's work posted to a forum of presumably intellectually curious people gets a bit beat up.
George Bernard Shaw worked out a new alphabet too. Tolkien built a language for books about goblins and elves. I can't image how they would be greeted. But perhaps it's just a bad day or hour or something.
I think generally we should strive for positive comments. But I have seen some pretty terrible Show HN’s. To be clear, they are the minority. But there are some that are basically “I will put drug dealing sex workers on the blockchain using Tor” or similar where I think it is warranted to point out that it’s a terrible idea or terrible execution or both.
A problem with HN is that the rules essentially incentivise this behaviour: if you simply think something is good, you cannot say so until someone else said it was not. “This is great “ is not a good comment. So people upvote instead. But scores are hidden , so you can’t see that from the outside.
you can see this effect in action on your own posts : 10 upvotes, 1 critique. From the outside however, it just looks like “HN is at it again!”
Sure, people could say “this is good because X”, just as they should with critique , anyway. But that’s not how we think : it’s easier to motivate why you don’t like something than why you do.
I think HN is much more positive than it seems. We just need to find a way to stop hiding it.
The website states “was created thousands of years ago, and is optimized for writing, not reading. About time for an update, no? Dotsies is optimized for reading.”, and other claims on the website, and people are not instantly convinced just because it’s free.
Positivity is fine, but blind positivity is not helpful for anyone..
> Positivity is fine, but blind positivity is not helpful for anyone..
I'm with you on that one, but at this point I'd even take blind positivity over the usual constant blind negativity to anything new on HN. Maybe that's the actual remedy to that, a bit of blind positivity in every thread to counteract the immediate all-around pessimism.
That’s fair. I guess the karma system makes for more combative comments, or comments that more align with the status quo. Like https://forums.bunsenlabs.org/viewforum.php?id=13 seems to be welcoming a lot more discussion. I can only give that as an example because it was the last non karma forum I’ve visited in ages, and even that was a long time ago.
To be fair, you could use your exact post to argue in favor of Juicero, uBeam or Theranos.
More in-depth, yes, a lot of posts here are criticizing - but it seems to me, a lot of it is constructive criticism from someone who actually did find the idea interesting.
I personally would find a HN where each post is just a variant of "woah that's super cool!" a lot less interesting.
Notably differently, Dotsies is not heavily funded, nor does it want you to pay several digits for access. It’s just a fun experiment. It should not be held to the same standards of usefulness as a thousand-dollar juicer.
Creator of Dotsies here. I've been using Dotsies to read books since 2012 (maybe around 100). At first on a Kindle and then on a Kobo after Amazon stopped letting you add your own fonts. I reached my normal reading speed a few years ago, and now reading it is fairly second nature. I now prefer it to normal fonts, and have the feeling that my eyes are racing unnecessarily from side to side when I use them. With normal fonts I'm also less likely to "seek" to different parts of the page, and thus feel a bit handicapped with them.
> the brain uses a hierarchy of constructs like circles, lines, corners, etc
It's a little counter intuitive if you haven't tried it, but the shape-seeking aspects of our visual systems do come into play with grids of dots. As others have said below, the words jump out at you as shapes. Sort of in the way that letters normally do. Just to approximate it, consider: :..'. But you may have to look at words a few thousand times before you get the full effect of this.
In general, the less visual noise and redundancy there is, the easier it is to recognize shapes. Consider looking at a pile of sand vs a pile of boulders, to get a sense of the extreme ends of this spectrum. Pile of sand: high complexity, harder to make out visual shapes. Pile of boulders: low complexity, easy to make out shapes. It is possible to map each letter to unique arrangements of 1000 disordered grains of sand, but clearly a few boulders per letter would be better. Consider that it takes about 20 pixels to represent each Latin letter. We scrapped reading fancy calligraphy because they would take more like 40px or so pixels to represent. And the extra flourishes really just added superfluity to the page. Only that was difficult for us to see, since we were so used to the calligraphy as the norm. I think it's naive to think the current state is the apex. Let's get rid of the remaining flourishes! Calligraphy is sand. Latin is marbles (or spaghetti might be apter). Binary grids are boulders. Just stare at the equivalent paragraphs under "How much better is it?" on dotsies.org for a while and you'll get a sense of this. (Especially if you can do so while imagining you're unfamiliar with Latin letters.)
I think I made a joke in 2012 about cave-men waving their spears at the stone-age version of hacker news claiming their bison heads and triangles couldn't possibly be improved on :).
Just wanted to give a quick summary. Will reply to other comments directly.
Btw I'm also the creator of Xiki, which has been discussed on HN as well, and has a stronger claim to originality/practicality. I did 2 kickstarters for it, which didn't get much hacker news attention, much to my disappointment. I'm doing a third soon and am hoping for more interest/abuse :). My new version will switch to using the markdown format, and will hopefully revolutionize the command line.
you should really cite your own personal experience on the page. if i had seen that you have read 100 books in dotsies i would have taken it much more seriously. all this other stuff is kind of post hoc rationalization. if dotsies didn't end up being easier to read it would all be irrelevant. i really think you underestimate the complexity of reasons (or lack thereof) why dotsies works (or doesn't)
This is amazing! You inspired me to get Dotsies on my Kindle. I've been wanting to read Cixin's "The Three-Body Problem", so this is the perfect opportunity.
I'm also packaging the fonts up for Void Linux but was unable to find any licensing information. Do you have dotsies under Creative Commons? Did I just miss something?
That still doesn’t explain very much. Your answer to “This has been done before.” is ‘SUPER. COOL.’, so what was invented? Could you provide some other things that are on your super cool list or some synonyms?
This is interesting for a lot of reasons. It’s almost like a compression scheme for text as opposed to an image. The question is, is it lossless or lossy? What is the Shannon entropy for this set of symbols?
Well I guess I am more used to the physicist’s definition of entropy, which can be described as “the higher the entropy, the less we care about the details.”
Very possibly, but, as far as I know, in the information-theoretic sense (which is what we mean when we talk about alphabets and conflating one letter for another) it's "high entropy" = "low redundancy".
This doesn't compress the text any more than reducing font size does. The Shannon entropy is about the frequency with which each symbol occurs, not about what the symbols look like. This is just an alternate font and does not actually change symbol frequency, so the Shannon entropy will be the same as any other font.
It's lossless unless you consider the possibility of symbol confusion, as it's pretty easy to produce strings that appear identical but are different, like how O and 0 are easily confused in many fonts.
you could "solve" the same "problem" by making font smaller. but its annoying to read. oh whoops, thats what actually matters.
this page is like the purest expression a cerebral cortex thinking its the only thing that exists. i like it though, because it accidentally makes some good points about the human condition. one is that top down designed shit like this (and esperanto) often sucks incredibly hard, and i think the reason is that the "problems" they fix aren't problems at all.
another is that its hilarious to claim to solve a problem by asking someone to first relearn to read. "just spend 1000 hours learning this, then it will be easy!".
another is how easily people seem to oversimplify what is happening when you do something. someone recently told me that an upright bass is just a "dumb guitar" because its tuned the same and has fewer strings. sure, if you reduce the playing of an instrument to the production of a certain pitch, and that alone (even then... no, but i digress...) but of course thats really dumb. as is the case with "dotsies". you've decided that our visual software is agnostic, and we can rework the written word to whatever suits your optimization criteria, which seem to be that you can fit more letters in a small space, and that words have a shape-like gestalt.
even if it were true that these two criteria are paramount, making a thin font of english letters would work way better. but i see no evidence or justification for those criteria being important. its not that you cant improve on things, but how do you know whats really difficult about reading? did you do research? did you study the literature? seems like there are lots of more obvious problems you could try to solve. this is all getting a little ranty, and i risk sounding stupid if it turns out you are coming from a place of knowledge. so if you are, id be curious to know (and why isn't that on the page?)
This reminds me of the old "real programmers use binary" meme, and how I experimented with bitmap image files by typing text into them and looking at the resulting patterns.
I wonder if I'm the only one who would find it easier to read if it was just ASCII printed in binary instead of its own symbol-set. ASCII has some patterns which make it easy to memorise, and it's something a lot of developers have indeed memorised already. That would make a great "1337 hax0r" font.
...and why not go the whole hog and make it 8 bits, so you can encode all of Unicode (as UTF-8)...
On a barely more productive note, I am pretty sceptical that an "optimized" alphabet glyph would be one dot for a b c d e, two adjacent dots for f g h i, etc. How likely is it that alphabetical order also happens to be the best dot pattern?
Interesting, similar to how the korean alphabet works; letters are composed into syllabic units. “Lol,” for example, is 롤, which is a combination of ㄹ + ㅗ + ㄹ. It’s easy to see how they match up with their latin counterparts.
Very interesting. It's definitely more compact but I would challenge the more readable claim.
If you think about how the human vision system is constructed (similar to convnets in deep learning), the brain uses a hierarchy of constructs like circles, lines, corners, etc. Recognizing letters requires the triggering of multiple layers of these constructs.
A dot, and another dot 1 pixel higher or lower is easy to get conflated. There's going to be a high degree of uncertainty. We have a hard enough time distinguishing '1' and 'l', and 'O' and '0'. This makes it so every letter takes on that trait. We usually must rely on context of the surrounding letters and words to fill in the missing information.
I think the concept of this idea is valid, but we need more than moving pixels up and down to distinguish letters. If instead a form existed that mixed various small primitive shapes it might work.
I'm not an expert on the vision system but I would think, sticking to at most 2 different locations instead of 5, and mixing in vertical lines, horizontal lines, diagonals, corners, and circles might be work exploring.
There's also color but lots of people are color blind.
> I'm not an expert on the vision system but I would think, sticking to at most 2 different locations instead of 5, and mixing in vertical lines, horizontal lines, diagonals, corners, and circles might be work exploring.
What you proposed here turns out very similar to the elements of Korean Characters.
As an East Asian who is native to Chinese Characters, two-dimensional symbols are always more appealing to me. I started
to learn English 25 years ago while Japanese 10, but I read Japanese much more faster than English even without any Kanji.
I'm impressed, If you remove the Kanji from Japanese, my reading speed goes down significantly.
Though, as a native English speaker of course my English (alphabet) reading speed is faster than my Japanese (Hiragana/Katana) speed.
What I do find very interesting though is that given I know all/most of the words in a Chinese dialogue, I have noticed that my Chinese reading speed is close to on-par with my English reading speed. I find Chinese very efficient to read/parse. I think there are many factors that contribute to this such as well defined word roots, similar sized characters, relatively small number of radicals (larger than alphabet of course), very simple/clean grammar.
The more I study, learn, and use Chinese, I really wish Chinese was the international language. It's power and elegance always amazes me. Even more so as I get further into Chinese literature, poetry, culture, humor, etc.
I've studied Japanese/Chinese both for a long time (10-15 years). Nowadays, I use Chinese everyday as that's the language my wife and I use.
As a pure alphabet goes, I really like the Korean implementation. I was able to mostly learn and retain it with only 1-2 hours of study.
Interesting, thanks for sharing your perspective as a native English speaker and a Chinese learner.
> I really wish Chinese was the international language. It's power and elegance always amazes me. Even more so as I get further into Chinese literature, poetry, culture, humor, etc.
It had been a dominating language/writing system for part of Asia, and we can still find its legacy in Vietnamese, Korean, Japan.
But I think "legacy" is the keyword here. As a software engineer, the Chinese writing system is not in a ideal shape at least for me. There used to be systematic ways/principles to create characters, but there has never been any sustainable approach to "maintain" some of the internal consistency. It's like stacking workarounds one after another, and it becomes what we see today. Users can just accept it without any efficient way to modify or even refactor it, not mentioning the burden and effort to create just one beautiful font design.
I initially learned traditional Chinese characters (lots of Taiwanese friends), and the traditional variant that Japanese used. Even though initially I was stubborn and would claim that traditional characters are more beautiful or pure, I have to admit that simplified Chinese is definitely a major language improvement. My one exception is I still like to write 龍 instead of 龙 :)
I agree with you that there must be a newer/better way to construct a Chinese-like writing system that isn't quite a burden to both learn as well implement in computers (fonts, etc). I.e. Words could still be built up of logic word roots , similar to current system but more consistent/logical. More advanced words could be built of simpler words. Some of my favorite examples are words like "skull" that are in Chinese 骨头 (literally "head bone"). These sorts of constructions for me as a foreigner learning Chinese are amazing. I'm able to memorize so many Chinese technical/science terms that are a pain even for natives in English.
I sometimes have imagined training a neural network to generate "word vectors" that are actually just small 32x32 colored images, just to see what they would look like. Sort of like the word2vec generations, but using images as the input and the output would be a sort of viewable "character". I also imagine "hyper-intelligent" beings developing languages like this, only with higher dimensionality of input and not only images.
> I have to admit that simplified Chinese is definitely a major language improvement. My one exception is I still like to write 龍 instead of 龙 :)
Improvement or not? I argue that it depends. Once in an international situation, Mr. Xi wanted to speak a seldom-used and very old idiom in his draft, which was "通商宽农", roughly standing for the meaning that farmers (农) can live much better (宽) if economy is getting good. However, he spoke it as "通商宽衣"， which is not a established phrase but can still be reasoned as "get economy better and get undressed (宽衣)"。
As a reference, the traditional version of this idiom is "通商寬農"。"農" (farmer, agriculture) has a very different shape compared to "衣" (cloth, surface) in traditional Chinese, but obviously they looks similar when you become a important but nervous guy.
If we regard the Chinese language/writing system as a huge project, I would say that simplified Chinese was a hard fork without much testing but soon widely adopted, and its design was based on some novel but destructive compression algorithm without any peer review process.
> I sometimes have imagined training a neural network to generate "word vectors" that are actually just small 32x32 colored images, just to see what they would look like. ...
Cool idea! Sounds like some kind of esoteric language. Looking forward to it.
Haha that’s a great example. Those are definitely areas where the written language has room for improvement. I could also imagine myself making a similar mistake.
Regardless, I’ve always been impressed that the Chinese can make such a bold change as their written language yet here in the United States, I’m still arguing with people who don’t know how many feet are in a mile as to why we should switch to metric. :)
If I ever get around to making my “hyper language”, I’ll be sure to post or let you know. I imagine it will likely be “not-very-user-friendly”. haha
About the neural network part, not exactly what you're looking for but you might enjoy this blog post[1] by David Ha with a neural network "dreaming" fake kanjis.
Part of my motivation with Dotsies was to capture the compactness and visual distinctness of Asian characters, while still keeping what's good about alphabets. I'm guessing native Chinese speakers find it exhausting to have their eyes dart back and forth at a comparatively hectic speed to take in the same information when reading English?
Yep, that's Korean. Hangul groups all the letters in each syllable into a compact square, but the square can also be easily decomposed into individual alphabets when needed.
People who propose new writing systems would benefit from studying existing systems a bit more. The Latin alphabet is not the only game in town.
Hangul (the Korean writing system) came to mind immediately for me too. I can't read Korean, but the writing system is very well designed, being phonetic and compact. It would be great to have something like that for written English.
"
For example, in Year 1 that useless letter "c" would be dropped to be replased either by "k" or "s", and likewise "x" would no longer be part of the alphabet. The only kase in which "c" would be retained would be the "ch" formation, which will be dealt with later. Year 2 might reform "w" spelling, so that "which" and "one" would take the same konsonant, wile Year 3 might well abolish "y" replasing it with "i" and Iear 4 might fiks the "g/j" anomali wonse and for all.
Jenerally, then, the improvement would kontinue iear bai iear with Iear 5 doing awai with useless double konsonants, and Iears 6-12 or so modifaiing vowlz and the rimeining voist and unvoist konsonants. Bai Iear 15 or sou, it wud fainali bi posibl tu meik ius ov thi ridandant letez "c", "y" and "x" -- bai now jast a memori in the maindz ov ould doderez -- tu riplais "ch", "sh", and "th" rispektivli.
Fainali, xen, aafte sam 20 iers ov orxogrefkl riform, wi wud hev a lojikl, kohirnt speling in ius xrewawt xe Ingliy-spiking werld.
"
I doubt it. With very few minor correction this looks like phonetic transcription of what slavic (should hold at least for polish, czech and slovak) language native would write it down.
I believe german writing is also close to phonetical. That's why it works.
Grids of dots can be slightly morphed to resemble curves and slashes. Notice how some of the morphed versions of Dotsies appear visually similar to Asian scripts:
Interestingly, when you read Dotsies quickly and at a smaller font size, the scripted versions appear indistinguishable from the square (unmodified) versions. Because of this I didn't end up pursuing this aspect.
Dotsies would be trivial to learn if they are in digital form. Think of QR readers. I can't even imagine handwritten dotsies, they would probably be a mess.
Yeah, the whole claim "latin letters were optimized to be written by hand" is completely bollocks. The letters are also subject to constraints that are based in reading just as you described.
Also: compact representation doesn't lead to faster reading. This study [0] compares native readers of English and Chinese text; the latter being more information dense than the former on a per-pixel basis. As for reading speed? In words per minute, it's a wash: 380 vs. 390.
Dotsies words look kind of like letters. They have lines, corners, loops, etc.
> There's going to be a high degree of uncertainty. A dot, and another dot 1 pixel higher or lower is easy to get conflated
I initially thought that would be a big issue as well. Turns out in practice it almost never comes up when reading paragraphs. For headings or words on signs that could be more of an issue.
It's a damn shame this font isn't properly kerned, I put it in my term and immediately started throwing up. If the dotsies form a block like they do on the site, then the numbers/punctuation is overlapping and broken.
The terminal is tricky, due to the monospace aspect. It degrades readability when the spaces between the words are as skinny as the letters, but that's what a requirement of a monospace font. One (though not perfect) workaround is to make both the spaces and the letters a little fatter. Also making the numbers readable is a challenge with monospace. I've made numbers that are more adapted to the terminal, but they could use some improvement.
I believe it was an HN thread years ago where I first discovered Dotsies. I spent a very satisfying afternoon learning to read it and write it. I then labelled all sorts of cables and patch boards and such with it around my cubicle for fun. I only kept them up for a few days, but it was a blast learning to read something cryptic and cool looking at a glance with so little effort. If only for that idle distraction Dotsies is a worthy hack. I really think half the fun is in the teaching method.
It is fun. I actually used (and still use) the "Standard Galactic Alphabet" from the Commander Keen games every once in a while, which is a little easier to read and write, yet almost as baffling to non-insiders :)
There's also using multiple shades from dark to light. That might actually be easier to read if you can throw in a few light intensities are have less positional variants. 2 shades with 3 positions would be less error prone than 5 positions of only 1 shade.
I tried that, as well as with color variations. The problem that arises is when you view them at small or even medium font sizes, a black dot next to a red or gray dot just looks like a blurry black dot.
Why didn't the designer select dot positions based on letter frequency or letter pairing frequency in English? This could have yielded a font that would be more likely to make pretty geometric shapes. Even a simple improvement like vowels all having higher dot positions in the font, and consonants having lower positions, could be interesting. It seems that the dot positions are just based on letter position in the alphabet. There is probably a way to come up with dot positions that yield much nicer looking shapes.
My suspicion is that the way to improve legibility is to consider ways to make symbol images more robust to noise. In particular, if a symbol image were perturbed, would the context help you figure out what that symbol should be?
For each letter, we can build a histogram of contexts that letter occurs in for a sufficiently large corpus. An example is the letter "o", whose occurrences in the previous sentence have the following contexts: f_r (twice), t_g, c_n, _c, c_r.
Each letter's context histograms have some similarity with each other letter's. We want different letters that have similar contexts to have widely differing appearances, because if their appearances were similar, their contexts will be of less help to the reader in disambiguating. And letters with sufficiently different contexts may be allowed to be more similar in appearance (note e.g. the archaic long "s", which despite visual ſimilarity to "f" is easily distinguished from it by the different contexts it appears in).
That's a very interesting point and suggests that, counterintuitively, it would actually be bad for (for example) vowels to look similar to each other.
Another way to use language statistics would be to make the font variable-width (which may help with legibility) but make commonly-occurring letters thin.
Dvorak (for example) is an attempt to optimise for english text, so it is worth it at some level. I have no idea how applicable it is to other languages.
Hit the "random" link a few times and you'll get an idea of the possible variations. In the Dotsies pattern, vowels are at the tops and bottoms of words. The variations can look grainy, chunky, loopy, checker-board-ish (you'll see what I mean if you try it). The version I picked is a balance that leans toward being sparse and loopy, and is one of the many variations that has an intuitive letter ordering. Here's a decent one, but with a crappy ordering.
Surprisingly, none of them jump out as significantly better than the others (the overly dense ones being exceptions).
[from another comment]
> I am pretty sceptical that an "optimized" alphabet glyph would be one dot for a b c d e, two adjacent dots for f g h i, etc.
There are other mappings that looked marginally better to me, but I deemed picking one with a nice ordering was worth doing. A and E happen to be the 1st and 3rd most common letters, which works out if you want to lean toward the sparse end. Also O (the 4th), A, and E all stick to the top and bottom edges of words, so if you want your words to tend to have holes in the middle they are a good fit. Getting a decent mapping for the commonest letters seems to be key, but by all means, try the above link and post an alternative that you think is better.
To learn, read one word from the top paragraph and see if it "looks" like the bottom dotsies. Then gradually try to read the paragraph without looking at the next word.
I like your mapping just about as well as the Dotsies mapping. Slightly more speckled vs angular. I wish others who are finding fault with the mapping would go down this path as well. I think they would quickly realize that one mapping is (generally speaking) about as good as another, so you might as well pick one with a good order. That's why I created that page.
202 comments
[ 2.9 ms ] story [ 180 ms ] threadIt's a super fun idea, though. I would 100% put this in the background, or even the foreground, of some future scifi scape.
You're going to need damn good eyes to read Dotsies at anything like 13px, and I suspect even the sharp-eyed are going to misread a lot. Size Dotsies text up to the point that people of average vision can read it as reliably as a normal font at a reasonable size, and I don't think you're going to get much, if any, greater information density.
I'm reminded of Speedtalk, a language from Robert Heinlein's story "Gulf", which had every combination of phoneme and delivery (pitch, length, etc.) represent a word in the Basic English vocabulary. So, every spoken "word" in the language is actually a full sentence. But, it takes intense futuristic training just to be able to speak vaguely intelligibly in it, because all the normal imprecisions of speaking (much less speaking in a new language) take any tiny error from "You used 'record' as a noun, but with a long 'e' and the stress on the second syllable." to "You have a hovercraft, and it's full of what???"
Now imagine trying to follow speech in such a language if the speaker's out of breath, alarmed, surrounded by background noise, etc. Just not enough redundancy for reliable communication.
http://www.omniglot.com/language/phrases/hovercraft.htm
Dotsies seem to make a tradeoff where they optimize for higher information density at the cost of much higher error rate (or requirement for precision and accuracy if you will). I also definitely see your point about font sizes. Basically it’s a valiant attempt at a reimagined alphabet but I don’t think it actually achieves higher info density in the real world.
Can anyone who uses https://www.dyslexiefont.com/ comment on how hard it is to have it everywhere?
But Dotsies seems to be solving the wrong problem: using the fewest pixels. I think the right problem is being able to skim large amounts of text quickly, for which typical fonts are pretty good. You can find whether there are any numbers on a page, for instance, in well under a second.
No.
Once again, Silicon Valley "discovers" something that was already invented.
https://en.wikipedia.org/wiki/Braille
But of course, that's a specialized skill. To judge for yourself how conducive braille is to visual reading, when presented as black on white (or the opposite) instead of raised dots on a page, you can grab a braille font here:
https://www.hadley.edu/online_course_font.asp
They could maybe do better on legibility, or maybe it isn't that big a problem...the more I think on it, the more I get where the author was going (and feel they may have successfully reached it, or at least made strong progress)
"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."
Also:
"Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith."
Please review https://news.ycombinator.com/newsguidelines.html and follow the rules when posting here.
As to the originality, most computer science students have looked at a binary versions of ascii charts at one time or another. Turning them sideways and making a font out of it isn't discovering relativity. On the other hand, even Braille had its percursors (see your Wikipedia link), and Braille users don't reject it on that account, do they?
> Silicon Valley
You're assuming I'm from Silicon Valley. I am here now, but was actually in Ohio when I came up with Dotsies :)
It doesn't really make sense to optimize for screen space like this.
However, the teaching style along with a symbol alphabet that IS trying to solve for better legibility (without necessarily any allegiance to dyslexic fonts), could be pretty interesting.
In middle school when bored once I made my own symbolic alphabet, I found it really is extremely easy to learn a new symbol set for language you already know, so, changing away from the traditional symbols isn't even that massive of an endeavor.
I think they did do an OK job with using screen space well, but at the cost of legibility to a degree. That said, the more I think on it, maybe the author did too. Sure, individual letters aren't as legible, but the point is almost more that words become single characters, more like Mandarin. If those characters are legible, than it doesn't matter that the individual letters are not. If you don't know a word, you're already going to be individually inspecting the letters. (of course, it will still be mildly more difficult with dotsies)
At the same time it would be interesting to see it implemented based on letter frequency rather than order in the alphabet.
It was great fun to read!
Optimizing for legibility as opposed to screen space is a distinction that isn't very meaningful in practice. Because if words aren't legible enough, people increase the font size until they are. If something is extremely legible (like latin letters at a 40pt font size) people will naturally tend to decrease the size. So screen space and legibility are largely different labels for the same concept. With some wiggle room but not that much.
[cite needed][original research?]
Automobile
汽车
Reading chinese is basically the human version of attempting to read a .zip file.
This came up in the Robinhood checking thread a few days ago. "Do you really think that you're going to outsmart a financial institution that employs people like you (but smarter) because you happen to be better at math than your facebook friends who studied polysci?" The hubris of people in tech who think they're the only people who notice things is getting really tiresome.
Having said that, I think that's a slippery slope of a stance to take, because it stifles creativity and novel solutions, which history shows often come up when they're not looked for. But I think that's kind of the sentiment of embarrassment they were expressing: "groan, another techie is going to disrupt something that doesn't need disrupting."
Edit: To the people responding to this (and downvoting): this person asked for an explanation for why someone would feel embarrassment at this. I was trying to give what I thought was a reasonable explanation. I don't actually feel embarrassed by this, I was just trying to understand and articulate someone else's feelings on this. I'm sorry if the explanation didn't do it for you.
Ya know, the marquee sign does say Hacker News. If this seems like silly "tech people" nonsense to you, maybe you're not the intended audience.
I don't see a reason for the negativity, because I don't see a single disadvantage in trying. If the author wants to waste their time, let them.
Given how different some international alphabets are (Russian, Hebrew, Katakana) it's not self-evident that the conventional Latin glyphs are optimal.
I don't know all the fancy names for logical fallacies but this seems to be deliberately misrepresenting the argument. The website clearly states that the problem with the Latin alphabet is that it has to be able to be written efficiently as well as read. Your trillions of humans over a few thousand years have had to both read and write this alphabet. In the past few hundred years we've had technology that allows someone who made a huge investment to produce massive amounts of text without writing, and in the past few decades the ability to produce text without writing has become completely ubiquitous. Does it still make sense for this alphabet to optimize for hand writing? Are you arguing that the constraints that made the Latin (or any handwritten) alphabet have remained unchanged over thousands of years? They want to redesign an alphabet to fit modern constraints. We consume a lot of text on devices that need to have screens that fit in our pockets. Perhaps a two-thousand year old alphabet isn't still the optimal solution?
I also don't think the authors necessarily expect to disrupt anything. They just made something cool and put it on the internet. I'm happy people like this exist.
This isn't a unique aspiration of tech. Folks have been trying to achieve better othographies than the one we grafted on as a colonizer's comprimise for a long time. Projects like Shavian [0][1], Read Script (Quickscript) [2], and Deseret [3] come to mind as prior art in this space that predate the invention of the modern tech community and its current perceived class distinction.
I'd argue it's the supreme arrogance to assume that in a modern world populated with people who's average intelligence, nutrition and education was only accessible to the richest and most powerful 0.001% 200 years ago that NO progress can be made on a common linguistic platform. The classical alphabet we use has ALSO been revised and has literally hundreds of variants.
On a personal note, I don't at all regret the decision to learn to read and write in Shavian. Having a semi-private notation with which to take notes, form signatures, etc. The only thing I am sad about is that so few people in this community (and others) decided to put in the modest effort required to do the same. If I do occasionally offer thoughts in ·𐑖𐑱𐑝𐑾𐑯, 𐑐𐑰𐑐𐑤 𐑸 𐑩𐑯𐑜𐑮𐑰, 𐑮𐑧𐑕𐑧𐑯𐑑𐑓𐑳𐑤 𐑰𐑝𐑧𐑯! 𐑤𐑲𐑒 𐑞 𐑧𐑒𐑕𐑦𐑕𐑑𐑧𐑯𐑕 𐑝 𐑩𐑤𐑑𐑺𐑯𐑩𐑑𐑦𐑝 𐑹𐑔𐑪𐑜𐑮𐑩𐑓𐑰 𐑦𐑟 𐑩𐑯 𐑦𐑯𐑕𐑳𐑤𐑑. 𐑲 𐑛𐑪𐑯𐑑 𐑜𐑧𐑑 𐑦𐑑.
I'm debating if dotsies would be valuable to learn. My biggest complaint about it is that it's a reproduction of our current alphabet, and that means you'll end up with a lot of our weird di-and-trigraphs. If you're going to to the trouble of a con-script, why not use a phonetic base?
[0]: 𐑮𐑰𐑛 𐑥𐑹 𐑣𐑽: https://en.wikipedia.org/wiki/Shavian_alphabet
[1]: 𐑢𐑦𐑗 𐑲𐑥 𐑒𐑳𐑥𐑓𐑹𐑑𐑩𐑚𐑤 𐑢𐑦𐑞, 𐑕𐑴 𐑮𐑧𐑤𐑰𐑕𐑛 𐑱 ·𐑤𐑦𐑯𐑳𐑒𐑟 𐑒𐑰𐑥𐑩𐑐 𐑣𐑽: https://gitlab.com/endomain/xkb-shavian-variant
[2]: https://www.omniglot.com/pdfs/quikscriptmanual.pdf
[3]: https://en.wikipedia.org/wiki/Deseret_alphabet
Without knowing it, you've taken a weirdly extreme and limited perspective - and deprived yourself of anything remotely close to a complete understanding.
I assume you're probably proud of your "optimism" or something here, but it's absolutely silly.
I think, real hackers lose interest in monoalphabetic substitution ciphers by junior high.
Take a look at the shavian alphabet for another interesting take on written text.
> This is useless.
Who cares? It's super cool!
> This is not well-suited for use case X.
Did you read the page? It's so great that someone thought of this and did it.
> This has been done before.
SUPER. COOL.
> It's hard to read.
I don't know what you aren't getting about this.
> What if it implemented improvement X?
I don't know, I didn't write it. Maybe the author would appreciate the suggestion!
Mean comments allowable:
George Bernard Shaw worked out a new alphabet too. Tolkien built a language for books about goblins and elves. I can't image how they would be greeted. But perhaps it's just a bad day or hour or something.
you can see this effect in action on your own posts : 10 upvotes, 1 critique. From the outside however, it just looks like “HN is at it again!”
Sure, people could say “this is good because X”, just as they should with critique , anyway. But that’s not how we think : it’s easier to motivate why you don’t like something than why you do.
I think HN is much more positive than it seems. We just need to find a way to stop hiding it.
Positivity is fine, but blind positivity is not helpful for anyone..
I'm with you on that one, but at this point I'd even take blind positivity over the usual constant blind negativity to anything new on HN. Maybe that's the actual remedy to that, a bit of blind positivity in every thread to counteract the immediate all-around pessimism.
More in-depth, yes, a lot of posts here are criticizing - but it seems to me, a lot of it is constructive criticism from someone who actually did find the idea interesting.
I personally would find a HN where each post is just a variant of "woah that's super cool!" a lot less interesting.
Creator of Dotsies here. I've been using Dotsies to read books since 2012 (maybe around 100). At first on a Kindle and then on a Kobo after Amazon stopped letting you add your own fonts. I reached my normal reading speed a few years ago, and now reading it is fairly second nature. I now prefer it to normal fonts, and have the feeling that my eyes are racing unnecessarily from side to side when I use them. With normal fonts I'm also less likely to "seek" to different parts of the page, and thus feel a bit handicapped with them.
> the brain uses a hierarchy of constructs like circles, lines, corners, etc
It's a little counter intuitive if you haven't tried it, but the shape-seeking aspects of our visual systems do come into play with grids of dots. As others have said below, the words jump out at you as shapes. Sort of in the way that letters normally do. Just to approximate it, consider: :..'. But you may have to look at words a few thousand times before you get the full effect of this.
In general, the less visual noise and redundancy there is, the easier it is to recognize shapes. Consider looking at a pile of sand vs a pile of boulders, to get a sense of the extreme ends of this spectrum. Pile of sand: high complexity, harder to make out visual shapes. Pile of boulders: low complexity, easy to make out shapes. It is possible to map each letter to unique arrangements of 1000 disordered grains of sand, but clearly a few boulders per letter would be better. Consider that it takes about 20 pixels to represent each Latin letter. We scrapped reading fancy calligraphy because they would take more like 40px or so pixels to represent. And the extra flourishes really just added superfluity to the page. Only that was difficult for us to see, since we were so used to the calligraphy as the norm. I think it's naive to think the current state is the apex. Let's get rid of the remaining flourishes! Calligraphy is sand. Latin is marbles (or spaghetti might be apter). Binary grids are boulders. Just stare at the equivalent paragraphs under "How much better is it?" on dotsies.org for a while and you'll get a sense of this. (Especially if you can do so while imagining you're unfamiliar with Latin letters.)
I think I made a joke in 2012 about cave-men waving their spears at the stone-age version of hacker news claiming their bison heads and triangles couldn't possibly be improved on :).
Just wanted to give a quick summary. Will reply to other comments directly.
Btw I'm also the creator of Xiki, which has been discussed on HN as well, and has a stronger claim to originality/practicality. I did 2 kickstarters for it, which didn't get much hacker news attention, much to my disappointment. I'm doing a third soon and am hoping for more interest/abuse :). My new version will switch to using the markdown format, and will hopefully revolutionize the command line.
I'm also packaging the fonts up for Void Linux but was unable to find any licensing information. Do you have dotsies under Creative Commons? Did I just miss something?
How so? The entropy looks (dangerously) low to me.
"aaaaaaaaaaaaaaaaaaaaaaaa" is low entropy, "JkjgpUBn74AREExy" is high.
It's lossless unless you consider the possibility of symbol confusion, as it's pretty easy to produce strings that appear identical but are different, like how O and 0 are easily confused in many fonts.
this page is like the purest expression a cerebral cortex thinking its the only thing that exists. i like it though, because it accidentally makes some good points about the human condition. one is that top down designed shit like this (and esperanto) often sucks incredibly hard, and i think the reason is that the "problems" they fix aren't problems at all.
another is that its hilarious to claim to solve a problem by asking someone to first relearn to read. "just spend 1000 hours learning this, then it will be easy!".
another is how easily people seem to oversimplify what is happening when you do something. someone recently told me that an upright bass is just a "dumb guitar" because its tuned the same and has fewer strings. sure, if you reduce the playing of an instrument to the production of a certain pitch, and that alone (even then... no, but i digress...) but of course thats really dumb. as is the case with "dotsies". you've decided that our visual software is agnostic, and we can rework the written word to whatever suits your optimization criteria, which seem to be that you can fit more letters in a small space, and that words have a shape-like gestalt.
even if it were true that these two criteria are paramount, making a thin font of english letters would work way better. but i see no evidence or justification for those criteria being important. its not that you cant improve on things, but how do you know whats really difficult about reading? did you do research? did you study the literature? seems like there are lots of more obvious problems you could try to solve. this is all getting a little ranty, and i risk sounding stupid if it turns out you are coming from a place of knowledge. so if you are, id be curious to know (and why isn't that on the page?)
I wonder if I'm the only one who would find it easier to read if it was just ASCII printed in binary instead of its own symbol-set. ASCII has some patterns which make it easy to memorise, and it's something a lot of developers have indeed memorised already. That would make a great "1337 hax0r" font.
...and why not go the whole hog and make it 8 bits, so you can encode all of Unicode (as UTF-8)...
If you think about how the human vision system is constructed (similar to convnets in deep learning), the brain uses a hierarchy of constructs like circles, lines, corners, etc. Recognizing letters requires the triggering of multiple layers of these constructs.
A dot, and another dot 1 pixel higher or lower is easy to get conflated. There's going to be a high degree of uncertainty. We have a hard enough time distinguishing '1' and 'l', and 'O' and '0'. This makes it so every letter takes on that trait. We usually must rely on context of the surrounding letters and words to fill in the missing information.
I think the concept of this idea is valid, but we need more than moving pixels up and down to distinguish letters. If instead a form existed that mixed various small primitive shapes it might work.
I'm not an expert on the vision system but I would think, sticking to at most 2 different locations instead of 5, and mixing in vertical lines, horizontal lines, diagonals, corners, and circles might be work exploring.
There's also color but lots of people are color blind.
What you proposed here turns out very similar to the elements of Korean Characters.
As an East Asian who is native to Chinese Characters, two-dimensional symbols are always more appealing to me. I started to learn English 25 years ago while Japanese 10, but I read Japanese much more faster than English even without any Kanji.
Though, as a native English speaker of course my English (alphabet) reading speed is faster than my Japanese (Hiragana/Katana) speed.
What I do find very interesting though is that given I know all/most of the words in a Chinese dialogue, I have noticed that my Chinese reading speed is close to on-par with my English reading speed. I find Chinese very efficient to read/parse. I think there are many factors that contribute to this such as well defined word roots, similar sized characters, relatively small number of radicals (larger than alphabet of course), very simple/clean grammar. The more I study, learn, and use Chinese, I really wish Chinese was the international language. It's power and elegance always amazes me. Even more so as I get further into Chinese literature, poetry, culture, humor, etc.
I've studied Japanese/Chinese both for a long time (10-15 years). Nowadays, I use Chinese everyday as that's the language my wife and I use.
As a pure alphabet goes, I really like the Korean implementation. I was able to mostly learn and retain it with only 1-2 hours of study.
> I really wish Chinese was the international language. It's power and elegance always amazes me. Even more so as I get further into Chinese literature, poetry, culture, humor, etc.
It had been a dominating language/writing system for part of Asia, and we can still find its legacy in Vietnamese, Korean, Japan.
But I think "legacy" is the keyword here. As a software engineer, the Chinese writing system is not in a ideal shape at least for me. There used to be systematic ways/principles to create characters, but there has never been any sustainable approach to "maintain" some of the internal consistency. It's like stacking workarounds one after another, and it becomes what we see today. Users can just accept it without any efficient way to modify or even refactor it, not mentioning the burden and effort to create just one beautiful font design.
I initially learned traditional Chinese characters (lots of Taiwanese friends), and the traditional variant that Japanese used. Even though initially I was stubborn and would claim that traditional characters are more beautiful or pure, I have to admit that simplified Chinese is definitely a major language improvement. My one exception is I still like to write 龍 instead of 龙 :)
I agree with you that there must be a newer/better way to construct a Chinese-like writing system that isn't quite a burden to both learn as well implement in computers (fonts, etc). I.e. Words could still be built up of logic word roots , similar to current system but more consistent/logical. More advanced words could be built of simpler words. Some of my favorite examples are words like "skull" that are in Chinese 骨头 (literally "head bone"). These sorts of constructions for me as a foreigner learning Chinese are amazing. I'm able to memorize so many Chinese technical/science terms that are a pain even for natives in English.
I sometimes have imagined training a neural network to generate "word vectors" that are actually just small 32x32 colored images, just to see what they would look like. Sort of like the word2vec generations, but using images as the input and the output would be a sort of viewable "character". I also imagine "hyper-intelligent" beings developing languages like this, only with higher dimensionality of input and not only images.
Improvement or not? I argue that it depends. Once in an international situation, Mr. Xi wanted to speak a seldom-used and very old idiom in his draft, which was "通商宽农", roughly standing for the meaning that farmers (农) can live much better (宽) if economy is getting good. However, he spoke it as "通商宽衣"， which is not a established phrase but can still be reasoned as "get economy better and get undressed (宽衣)"。
As a reference, the traditional version of this idiom is "通商寬農"。"農" (farmer, agriculture) has a very different shape compared to "衣" (cloth, surface) in traditional Chinese, but obviously they looks similar when you become a important but nervous guy.
If we regard the Chinese language/writing system as a huge project, I would say that simplified Chinese was a hard fork without much testing but soon widely adopted, and its design was based on some novel but destructive compression algorithm without any peer review process.
> I sometimes have imagined training a neural network to generate "word vectors" that are actually just small 32x32 colored images, just to see what they would look like. ...
Cool idea! Sounds like some kind of esoteric language. Looking forward to it.
Regardless, I’ve always been impressed that the Chinese can make such a bold change as their written language yet here in the United States, I’m still arguing with people who don’t know how many feet are in a mile as to why we should switch to metric. :)
If I ever get around to making my “hyper language”, I’ll be sure to post or let you know. I imagine it will likely be “not-very-user-friendly”. haha
[1] http://otoro.net/kanji/
People who propose new writing systems would benefit from studying existing systems a bit more. The Latin alphabet is not the only game in town.
" For example, in Year 1 that useless letter "c" would be dropped to be replased either by "k" or "s", and likewise "x" would no longer be part of the alphabet. The only kase in which "c" would be retained would be the "ch" formation, which will be dealt with later. Year 2 might reform "w" spelling, so that "which" and "one" would take the same konsonant, wile Year 3 might well abolish "y" replasing it with "i" and Iear 4 might fiks the "g/j" anomali wonse and for all.
Jenerally, then, the improvement would kontinue iear bai iear with Iear 5 doing awai with useless double konsonants, and Iears 6-12 or so modifaiing vowlz and the rimeining voist and unvoist konsonants. Bai Iear 15 or sou, it wud fainali bi posibl tu meik ius ov thi ridandant letez "c", "y" and "x" -- bai now jast a memori in the maindz ov ould doderez -- tu riplais "ch", "sh", and "th" rispektivli.
Fainali, xen, aafte sam 20 iers ov orxogrefkl riform, wi wud hev a lojikl, kohirnt speling in ius xrewawt xe Ingliy-spiking werld. "
I believe german writing is also close to phonetical. That's why it works.
http://dotsies.org/script.html
Interestingly, when you read Dotsies quickly and at a smaller font size, the scripted versions appear indistinguishable from the square (unmodified) versions. Because of this I didn't end up pursuing this aspect.
[0]: https://link.springer.com/content/pdf/10.3758/BF03204913.pdf
> There's going to be a high degree of uncertainty. A dot, and another dot 1 pixel higher or lower is easy to get conflated
I initially thought that would be a big issue as well. Turns out in practice it almost never comes up when reading paragraphs. For headings or words on signs that could be more of an issue.
I find shavian particularly interesting since it is based on phonetic pronuncation with which I've had many issues and embarrassments!
https://www.omniglot.com/writing/shavian.htm
https://www.omniglot.com/conscripts/dotsies.htm
For each letter, we can build a histogram of contexts that letter occurs in for a sufficiently large corpus. An example is the letter "o", whose occurrences in the previous sentence have the following contexts: f_r (twice), t_g, c_n, _c, c_r.
Each letter's context histograms have some similarity with each other letter's. We want different letters that have similar contexts to have widely differing appearances, because if their appearances were similar, their contexts will be of less help to the reader in disambiguating. And letters with sufficiently different contexts may be allowed to be more similar in appearance (note e.g. the archaic long "s", which despite visual ſimilarity to "f" is easily distinguished from it by the different contexts it appears in).
Another way to use language statistics would be to make the font variable-width (which may help with legibility) but make commonly-occurring letters thin.
I'm not convinced I could differentiate M N and O without context, having a known vowel to differentiate the heights you're looking at would help?
Try it yourself!
http://dotsies.org/design-your-own
Hit the "random" link a few times and you'll get an idea of the possible variations. In the Dotsies pattern, vowels are at the tops and bottoms of words. The variations can look grainy, chunky, loopy, checker-board-ish (you'll see what I mean if you try it). The version I picked is a balance that leans toward being sparse and loopy, and is one of the many variations that has an intuitive letter ordering. Here's a decent one, but with a crappy ordering.
http://dotsies.org/design-your-own/#comeyhangrxdtwqpsfubzljv...
Surprisingly, none of them jump out as significantly better than the others (the overly dense ones being exceptions).
[from another comment] > I am pretty sceptical that an "optimized" alphabet glyph would be one dot for a b c d e, two adjacent dots for f g h i, etc.
There are other mappings that looked marginally better to me, but I deemed picking one with a nice ordering was worth doing. A and E happen to be the 1st and 3rd most common letters, which works out if you want to lean toward the sparse end. Also O (the 4th), A, and E all stick to the top and bottom edges of words, so if you want your words to tend to have holes in the middle they are a good fit. Getting a decent mapping for the commonest letters seems to be key, but by all means, try the above link and post an alternative that you think is better.
http://dotsies.org/design-your-own/#iaeouvxnsfrcbqlzkjwygphm...
A couple of design values:
- vowels 1 dot seems like a good plan
- ending with s / y should be easily recognizable
- tall letters should be tall (tlbdk etc)
To learn, read one word from the top paragraph and see if it "looks" like the bottom dotsies. Then gradually try to read the paragraph without looking at the next word.
Open to improvements!