SoylentNews is people
SoylentNews
Submitted via IRC for Sulla
On a bright fall morning at Stanford, Tom Mullaney is telling me what's wrong with QWERTY keyboards. Mullaney is not a technologist, nor is he one of those Dvorak keyboard enthusiasts. He's a historian of modern China and we're perusing his exhibit of Chinese typewriters and keyboards, the curation of which has led Mullaney to the conclusion that China is rising ahead technologically while the West falls behind, clinging to its QWERTY keyboard.
Now this was and still is an unusual view because Chinese—with its 75,000 individual characters rather than an alphabet—had historically been the language considered incompatible with modern technology. How do you send a telegram or use a typewriter with all those characters? How do you even communicate with the modern world? If you're a Cambridge-educated classicist enamored with the Greeks, you might just conclude Chinese script is "archaic." Long live the alphabet.
But, Mullaney argues, the invention of the computer could turn China's enormous catalog of characters into an advantage.
His argument is [...] about our relationship to computers, not just as physical objects but as conduits to intangible software. Typing English on a QWERTY computer keyboard, he says, "is about the most basic rudimentary way you can use a keyboard." You press the "a" key and "a" appears on your screen. "It doesn't make use of a computer's processing power and memory and the cheapening thereof." Type "a" on a QWERTY keyboard hooked up to a Chinese computer, on the other hand, and the computer is off anticipating the next characters. Typing in Chinese requires mediation from a layer of software that is obvious to the user.
[...] The Chinese way of inputting text—the software-mediated way—will win out, says Mullaney. Actually, it's already won out. Our mobile phones now have predictive text and autocomplete. It took the constraint of mobile to get Westerners to realize the limits of the simple what-you-type-is-what-you-get keyboard. But even then, you could only get Americans to go so far.
Read more at The Atlantic.
(Score: 2) by AthanasiusKircher on Thursday August 01 2019, @03:36PM (1 child)
Agreed. Nothing annoys me more than people who try to have long extended discussions with me via SMS texts. Texting is convenient and fine for a quick question about what time to meet up or something, but I hate "typing" on a mobile device. If I'm doing more than 3 back-and-forth texting exchanges about something, I often just say, "I'll email you." It's faster for me to switch to a computer and pull up email than it is to waste time "typing" on a mobile device. (And yes, I've tried a number of different input methods.)
This is potentially a disadvantage, but not necessarily if learning algorithms can figure out your style and make increasingly better predictions for you over time. Yes, there needs to be a portable method that allows you to transfer this predictive pattern to another device if you want. And perhaps a setting that says "freeze options" so that the algorithm stops "training" and just always gives you the same options from the same input from that point forward. If you have that, I don't know that this objection is very serious, as your input method will improve as you work with the interface, not unlike how humans have had their "pet tools" for centuries -- a person can cut better and faster with "my knife" than with other people's, etc. Often tools and machines would be customized by their owners to work better for them. Why is this any different?
(Score: 3, Insightful) by jbruchon on Thursday August 01 2019, @10:15PM
Don't get me wrong, I'm not complaining about predictive input; it can be much faster than non-predictive input, but it becomes a double-edged sword, especially once you need to type anything differently than usual. For example, if I use my laptop to constantly type out blog posts about my snotty, ill-informed opinions on the internet (which I gleefully do, I assure you) but then I eventually need to write a formal grant proposal or a research paper for a university course, I'm going to use a very different style and way of wording things than what the predictions are now strongly weighted towards. There's a good chance that my favorite expletives and insults will be predicted inappropriately on a regular basis and I'll be forced to work around them in the prediction lists. This could end up actually slowing me down.
Of course, one of the constant problems with predictive input methods is that they not only may predict differently than expected and jolt you into stopping and being forced to concentrate on massaging the predictive inputs more cautiously, they'll also sometimes get all of the predictions wrong. At that point, you'll be forced to type most or all of the word(s) out anyway. In my experience, predictive keyboards are never ever faster than what a reasonably fast typist can produce.
The problem with Asian languages (particularly Chinese and Japanese) is that there are so many possible characters in the total character set that there's no practical way to have a keyboard with all of the possible characters. Languages with literally thousands of glyphs are wonderful for visual compactness, but they're absolute garbage once you aren't physically writing the characters out. They are hard to read at smaller sizes because they rely so heavily on fine detail to convey several shapes in one "box", they can't be typed in a practical way because there are just way too many glyphs to lay out, and they require 16+ bits per glyph to represent in a computer. It's faster to type kanji out as hiragana and then use a prediction engine to convert it, yes, but that's not because it's BETTER, it's because the only other ways possible to do it are very impractical. One of the beautiful things about the English language is that it centers around a fairly minimal set of easy-to-discriminate glyphs and the average English word length in common usage is roughly 5 characters. English is easy to type, fits on a hardware keyboard, and doesn't really need predictive techniques to speed up its input when the input device isn't trash (like a touchscreen.)
I'm just here to listen to the latest song about butts.