6 ms·

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 (s

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.

qlk1123 8y ago
> 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.
- KennyCason 8y ago
  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.
  - qlk1123 8y ago
    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.
    - KennyCason 8y ago
      Thanks for your perspective as well. 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.
- kijin 8y ago
  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.
- blueblimp 8y ago
  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.
  - jacobush 8y ago
    Like what Mark Twain suggested. https://msu.edu/~jerrymc/humor/spelling.html https://msu.edu/~jerrymc/humor/spelling.html " 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. "
    - krautsourced 8y ago
      Funnily enough, that's very close to how a non-English-speaking German would probably spell English words :)
      - idle_zealot 8y ago
        I believe that was the intended joke.
        YayamiOmate 8y ago
        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.
- lifthrasiir 8y ago
  I don't think Hangul-like writing systems are good fits for English however. Korean has a relatively simple syllable structure /C(G)VC/ and Hangul thoroughly exploits this trait. English does not; consonant clusters like "strike" are common, and every Hangul-like alternative to Latin alphabet for Korean was jarring to my eyes partly for that reason, e.g. [1]. It would be a bit more easier to work with word-based typography, i.e. logogram; refer to [2] and [3] for such attempts. [3] might be an interesting trade-off between logogram and phonetic alphabet. [1] http://www.omniglot.com/conscripts/engul.htm http://www.omniglot.com/conscripts/engul.htm [2] https://www.zompist.com/yingzi/yingzi.htm https://www.zompist.com/yingzi/yingzi.htm [3] http://www.omniglot.com/conscripts/swc.htm http://www.omniglot.com/conscripts/swc.htm
- trogdoro 8y ago
  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: http://dotsies.org/script.html 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.
punnerud 8y ago
Could we compare Dotsies learnability with "normal alphabeth" by checking which one is easiest to train with deep nevral network (NN)?
- visarga 8y ago
  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.
theon144 8y ago
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.
zamfi 8y ago
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. [0]: https://link.springer.com/content/pdf/10.3758/BF03204913.pdf https://link.springer.com/content/pdf/10.3758/BF03204913.pdf
trogdoro 8y ago
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.

qlk1123 8y ago

> 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.

KennyCason 8y ago
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.
- qlk1123 8y ago
  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.
  - KennyCason 8y ago
    Thanks for your perspective as well. 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.
kijin 8y ago
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.
blueblimp 8y ago
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.
- jacobush 8y ago
  Like what Mark Twain suggested. https://msu.edu/~jerrymc/humor/spelling.html https://msu.edu/~jerrymc/humor/spelling.html " 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. "
  - krautsourced 8y ago
    Funnily enough, that's very close to how a non-English-speaking German would probably spell English words :)
    - idle_zealot 8y ago
      I believe that was the intended joke.
      - YayamiOmate 8y ago
        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.
lifthrasiir 8y ago
I don't think Hangul-like writing systems are good fits for English however. Korean has a relatively simple syllable structure /C(G)VC/ and Hangul thoroughly exploits this trait. English does not; consonant clusters like "strike" are common, and every Hangul-like alternative to Latin alphabet for Korean was jarring to my eyes partly for that reason, e.g. [1]. It would be a bit more easier to work with word-based typography, i.e. logogram; refer to [2] and [3] for such attempts. [3] might be an interesting trade-off between logogram and phonetic alphabet. [1] http://www.omniglot.com/conscripts/engul.htm http://www.omniglot.com/conscripts/engul.htm [2] https://www.zompist.com/yingzi/yingzi.htm https://www.zompist.com/yingzi/yingzi.htm [3] http://www.omniglot.com/conscripts/swc.htm http://www.omniglot.com/conscripts/swc.htm
trogdoro 8y ago
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: http://dotsies.org/script.html 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.

punnerud 8y ago

Could we compare Dotsies learnability with "normal alphabeth" by checking which one is easiest to train with deep nevral network (NN)?

visarga 8y ago
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.

theon144 8y ago

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.

zamfi 8y ago

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. [0]: https://link.springer.com/content/pdf/10.3758/BF03204913.pdf https://link.springer.com/content/pdf/10.3758/BF03204913.pdf

trogdoro 8y ago

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.