Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Facebook is wrong, text is deathless

kottke.org

Post reply on HN
Print view View original

121–130 of 175 posts

Re: Facebook is wrong, text is deathless

#121

Post by x4m »

I think that interesting topic is "will we create programs as text for a long time?". Projection IDE seems to stall. Some time ago I was thinking about visual query language for databases (domain-specific) and composed this list of text advantages:

1. Version control systems (VSC) out of the box. Software developers have been using safe source code tracking for many decades. Since queries are the main tools for data analysts, they should be treated the same way, but the development of all VCS features in a custom query editor is not economically viable.

2. Portability. Text queries can be written even on a whiteboard and a notepad. One of the great advantages of text queries is that they are unambiguous: there are no hidden parts in a text query.

3. Detachable. Text queries can be run in different warehouses. They should not depend on identifiers, conditions, and environmental variables.

4. Fragmentation. An analyst can extract some feature from a query and partly pass it to his colleagues.

5. Embedding. Software developers easily can embed queries into autotests, side subsystems can embed query parts into their code or resources.

6. Specification. DSQL can be a part of a system’s applied programming interface (API) for third party systems if DSQL specification is precise enough.

Re: Facebook is wrong, text is deathless

#122

Post by trevyn »

enobrev wrote: post #118
hanniabu wrote: post #116

It'd be great if we could extract subtitles from video and put into a nice text format that's easy to read so it's skimmable. Then when you get to an area of interest, you should be able to click the place where you're interested and a video will pop up and start from there. Then when you've heard what you wanted to hear you can minimize the video and skim the text for the next area of interest and so on.

I imagine there could be an interesting means of moving the playhead along with your skimming, so when you highlight a block of text (as I do, habitually), the video would play through the selection and pause at the end. I'd love to use text as a video controller.

This is a great idea -- text is much faster to read, but watching a speaker is much more information-dense due to body language and intonation.

This would be a fun hack. Looks like the YouTube caption data is available via API: https://developers.google.com/youtube/v3/docs/captions/downl...

Re: Facebook is wrong, text is deathless

#123

Post by sandworm101 »

thaumasiotes wrote: post #119
sandworm101 wrote: post #76

Earlier quoted context omitted.

Those deep structure may be active, but reading was never in their design documentation. Our brains evolved to what they are today long before reading was a thing. The fact that we read so well is probably a hacked-together scheme tying together structures meant for pattern recognition and speech. I see reading much like swimming. None of us can swim without practice, as none of us can read. The scary thing is that e…

> Our brains evolved to what they are today long before reading was a thing. You could be on a poster for "evolution stops at the neck". Obviously, our brains did not evolve to what they are today long before reading was a thing, because that happened before today. There was no point at which brains stopped evolving.

>> There was no point at which brains stopped evolving.

Evolution operates on a different timescale. For all intents and purposes, our physical structure hasn't evolved since reading became widespread in the population perhaps a couple hundred years ago.

Re: Facebook is wrong, text is deathless

#126

Post by dexterdog »

jstimpfle wrote: post #99
dexterdog wrote: post #95

Earlier quoted context omitted.

That's why podcasts are so much better on 2-2.4x speed with a silence skipper. I get an effective 3x speed and as long as I'm paying attention I still get every word. It sounds like gibberish to most newcomers but if you slowly speed up you don't even notice the increments.

2.4x sounds incredible to me. I can't hear most lecturers at a speed faster than 1.6x.

Lectures may require a slower speed. I am talking about more general podcasts where there is more of a discussion going on. If the speaker is fast or if there is a wealth of numbers being recited you may need to slow it down. I don't have a single podcast that I run slower than 2x which yields an actual 2.6x with silence skipping.

Re: Facebook is wrong, text is deathless

#127

Post by thaumasiotes »

thomasfoster96 wrote: post #117
gumby wrote: post #87

Earlier quoted context omitted.

> My recent forays into Kanji suggest there's more to visual processing of text than what we permit ourselvesto learn due to phonic constraints associatedwith language acquisition The phonic connection to reading is overrated IMHO and this is why I wrote of a hack of the visual system. Kanji/Hanzi are so easy to read because over the millennia the ones that are hard to recognize at speed have bene dropped or tuned --…

> In fact I think the trend to teach kids English via "phonics" makes spelling harder; comment back if you want to discuss further. It took me a long time to realise the point of learning 'phonics' at school (they were called phonograms in Australia). While I can see why writing and reading without speaking makes phonics look useless, for children, all of their language is initially communicated through sound. Puttin…

> Of course, English uses 26 letter to represent 44 or so sounds, so as far as learning to read and write via phonics goes, English is one of the worst examples there is.

What you really want, for learning to read and write via phonics, is determinism -- you want someone to be able to write something if they know how it sounds, and to be able to read something if they see its written form. There's no need to have symbols and phonemes correspond one-to-one, and they generally don't even when people are devising their own orthography in a green field.

For example, using C for a generic consonant, we have the following regular spelling rules in English:

    aC -- TRAP vowel
    aCe -- FACE vowel
    eC -- DRESS vowel
    eCe -- FLEECE vowel
    iC -- KIT vowel
    iCe -- PRICE vowel
    oC -- LOT vowel
    oCe -- GOAT vowel
    uC -- STRUT vowel
    uCe -- GOOSE vowel
Voila, ten sounds in five symbols. The problem from a phonics perspective isn't that we don't have ten different vowel symbols for these, it's that we have other ways of writing the same vowels ("beat" is not written "bete") and other ways of reading words that appear to conform to these rules ("debris" does not rhyme with "priss", or even with "his").

Re: Facebook is wrong, text is deathless

#129

Post by nommm-nommm »

sandworm101 wrote: post #114
Tharkun wrote: post #113

Earlier quoted context omitted.

> structures meant for "Meant"? Nothing in your body or brain is "meant" for anything. There is no "meaning", or "intention", for starters. And the brain is nothing if not adaptable. If your brain is "meant" to be anything, it's versatile.

There are parts of the brain meant for specific tasks. It isn't a homogeneous mass of neurons. The bit that is working the eyes atm was designed (via evolution) to work the eyes. As was the bit that takes the message from the eyes and ups the heart rate when a lion appears in your field of view. There is certainly some flexibility, but the basic task layout is a layout, not a random allocation.

>The bit that is working the eyes atm was designed (via evolution) to work the eyes.

Evolution is a random process, there is no designing, there's no goals.

Re: Facebook is wrong, text is deathless

#130

Post by thaumasiotes »

gumby wrote: post #87
superobserver wrote: post #61

Earlier quoted context omitted.

My recent forays into Kanji suggest there's more to visual processing of text than what we permit ourselvesto learn due to phonic constraints associatedwith language acquisition, but there's practically nothing for me to go on other than look into how the deaf process textual representations independent of the kinesthetic and proprioception encodings of semantic features. Text certainly isn't dead, and I would rather…

> My recent forays into Kanji suggest there's more to visual processing of text than what we permit ourselvesto learn due to phonic constraints associatedwith language acquisition The phonic connection to reading is overrated IMHO and this is why I wrote of a hack of the visual system. Kanji/Hanzi are so easy to read because over the millennia the ones that are hard to recognize at speed have bene dropped or tuned --…

> Kanji/Hanzi are so easy to read because over the millennia the ones that are hard to recognize at speed have be[en] dropped or tuned

Um.... there are plenty of characters that would, at first glance, appear to be identical to each other. There are plenty more that people have difficulty identifying out-of-context, but not when they appear in context. Characters don't get dropped for being hard to recognize.

They are easier to read, assuming you already know them, because they encode significantly more information than a phonetically-based script does. That's a fairly straightforward example of more work up front allowing less work later, much as encoding wikipedia into a decompressor allows the compressed version of wikipedia to be very small.

Post reply on HN