Squirt.io – Readability Meets Spritz Speed Reading

[timtadh]

What I noticed using this to read a couple Ars Technica articles was it worked really well for short words. But, if a longer complicated word appeared, at say 600 wpm, I would miss it. It seems like an adaptive algorithm based on word length would improve the speed even more allowing it to go faster on short common words and slow down on longer unusual words.

Also, there is bug that sometimes causes two words to appear at once. Sometimes they are overlapping vertically which is basically impossible to read. Try reading http://pragdave.me/blog/2014/03/04/time-to-kill-agile/ to see what I mean.

[pkghost]

Hey timtadh! Spritz Inc's blog (http://www.spritzinc.com/blog/) has some interesting things to say about word recognition, primarily that it's related to a word's unique shape more than length. Words that are less than 4 characters long are actually harder to recognize due to their having more visual analogues than longer words, and words over 7 characters are start looking "long" more than any other shape characteristic. I've got some ideas about how to incorporate this into my app without doing a bunch of image processing, so, until then, keep your eyes peeled ;) Thanks for the feedback!

[eik3_de]

Slow down on long words would be very helpful, I just tried it with some german news text[1] at 600wpm: Worked great except for words that occured like

    Kinderbetreuungsmöglichkeiten
    US-Geheimdienstausschusses
    Bundestags-Innenausschuss
    Bundeswirtschaftsminister
    Bundesarbeitsministeriums

[1] http://www.deutschlandfunk.de/nachrichten.353.de.html

[jvanenk]

It's also doesn't work well for hyphenated words: 'step-by-step' and 'hastily-implemented' are flashed up all at once.

(Regardless of whether or not those are properly hyphenated.)

Edit: But this is still very cool.

[wpears]

That's what I did with Spree [1] (also posted below). Just a simple word.length*8 seemed to have quite a nice effect on readability.

1. https://chrome.google.com/webstore/detail/spree/aehoaolhojlm...

[seferphier]

Agreed. They need to slow down the speed for longer words. Otherwise, it would be perfect.

[nashequilibrium]

HN always likes the faster is better, I am a genius absorbing this much information quicker than you. Its been proven over and over that comprehension is way more important than speed of consumption. You guys fool yourselves into believing you can comprehend while reading at breakneck speeds, its way more fun & time saving to be able to imagine, driftaway in thought and understand what you are reading.

[shittyanalogy]

Not to mention that 1 word at a time makes it hard to:

    Repeat that word
    Scan back a few words
    Quickly see your position in the article
    Look at any sidenotes, footnotes, or relevant pictures
    Read with flow and not drum like rhythm
    Skip large sections of non-relevant material
    Back reference

[olalonde]

There isn't that much substance in most blog posts / articles I read online. Of course, I wouldn't use this to read "Artificial Intelligence: A Modern Approach".

[X4]

meh, I think you generalize a lot here. There are people who are very comfortably reading at 1000wpm and beyond and keep all of the story in the mind. Some people can even accurately repeat the sped read texts. There are more people who can do that, than you may think.

Personally I would NOT YET read a real book this way, for one reason: I cannot keep a person's name in memory well enough and would mix up roles and who said what. Even though I very rarely read romans (mostly publications) you can imagine how often you hear names in a text. But that's maybe just me, idk.

I guess you would feel embarrassed, if you couldn't like me keep persons names in association with their faces well enough. Example: 100-200 persons you meet every day and you remember 50names at most and that's only because you know them very well. You have to say 'hi' to all the others you just can't remember the name of. Luckily I don't mix up faces that much up anymore. Recently I met a girl whose name I simply couldn't recall although I've seen her 100s of times. Anyway she reminded me that I already have her in my phone-book, so yeah I'll try to keep it memorized.

[pkghost]

OP here. Totally agree. I would not use this to read a novel, but rather in the way that we already read articles on the web--skimming them.

I think there is a huge value to giving the reader a tradeoff between speed and comprehension, so long as the reader is conscious that they're making a choice and does so appropriately.

[andrewflnr]

Spritz is supposed to improve comprehension too. That's how it was pitched in the original article I read.

[EC1]

Honestly I don't have time to read, so I'm going to use this heavily. Long as I get the gist. If I miss a few things, I'll go back and quickly scan those few things I missed. I can finish like 10 articles on my commute no problem with this, versus just a few.

[philmcc]

I'm pretty sure this is just a big middle finger to Spritz. In the most hilarious way possible. Or am I misinterpreting the acknowledgments? ;)

[gohrt]

Yes. Squirt.io is engaging in the best kind of cultural warfare: Creating something valuable and sharing it with the world, undermining the profiteers who want to lock (other people's ideas) behind a paywall, who (attempt to) justify their claim with slick marketing brochures

[ghc]

I've never thought much about my reading speed before, but after trying this and finding the deafault setting a bit slow, I decided to compare my natural reading speed.

I was able to read the article I chose in about 85% of the time of the default 400WPM of the bookmarklet (I read the article first with a timer, and then reread with the bookmarklet), which would put me at 470WPM. With Squirt, I was only able to get up to 650WPM before it felt too uncomfortable.

I wonder if people really read one word at a time, especially when they're short words. If people do read more than one word at a time, I think a more intelligent approach might be necessary to really make the experience both comfortable and fast.

[pmichaud]

I tried using it on a github wiki page, and it broke because github wouldn't load the external javascript. So I copied the text and made a local html file. That didn't work because of these errors:

Failed to load resource: net::ERR_FILE_NOT_FOUND file://www.squirt.io/bm/font-awesome.css Failed to load resource: net::ERR_FILE_NOT_FOUND file://www.squirt.io/bm/squirt.css 62 Uncaught TypeError: Cannot read property 'style' of null squirt.js:10 48 Uncaught TypeError: Cannot read property 'style' of null

[Mizza]

I'll also shamelessly plug my open source implementation, OpenSpritz, which now has a healthy and vibrant community!

https://github.com/Miserlou/OpenSpritz

as well as the Android / Google Glass companion:

https://github.com/OnlyInAmerica/OpenSpritz-Android

[sitkack]

U should change your name. You are giving free advertising to something that doesn't need it.

http://en.wikipedia.org/wiki/Rapid_serial_visual_presentatio...

[encoderer]

Your bookmarklet never worked right for me. It wouldn't start. I'd select WPM and.. nothing. On latest chrome. Just FYI.

[Mojah]

This project may consider a name-change, as googling for this after it's gone from the HN homepage will trigger ... interesting results. Especially if you're trying to show this to a co-worker in the office.

Note to my boss: I'm sorry!

[ssully]

I believe that is part of the whole knock against spritz. In German, where some of the founding members of Spritz are from, spritz means a similar thing that squirt does in American culture.

[gohrt]

Welcome to humanity. Every verb is a euphemism for a sexual act.

[ta9876]

> on('squirt.again', startSquirt);

https://github.com/cameron/squirt/blob/gh-pages/bm/squirt.js...

priceless!

[jamiemchale]

Tweeted about it with hashtagging the name. Checked the hashtag. Deleted the Tweet.

[nazca]

Yeah, the name just sounds dirty. I changed the name of the bookmark...

[ZeGoggles]

Oh good lord. That is not a safe image search.

[devindotcom]

It's fun to see these ideas multiply and combine, but experts on reading comprehension seem to think RSVP didn't work even back in the 70s when it was being first tested:

http://www.nbcnews.com/tech/tech-news/not-so-fast-speed-read...

Comprehension was worse than skimming, apparently. I can see this as a thing to flash headlines/latest news to you but anything longer than a sentence or two... not so sure.

[rmyers]

I’m the creator of Acceleread, which is a reading trainer app for iOS. http://www.acceleread.com

Your point about comprehension is super important and often gets overlooked, so I wanted to add my 2 cents.

While the RSVP method seems to being used as a proxy for all things speed reading, it’s really only one part of the picture. RSVP is definitely faster (because eye movements take time) and particularly good for people with learning disabilities like dyslexia (because the tool isolate words), but there are trade-offs.

For one, saccades (the normal eye movements you make while reading) give your mind time to process the information. You need that. Secondly, RSVP tends to lead to repetition blindness, so you can miss repeated words, which changes your understanding. Also, people tend to zone out after a while of staring at one point. All of these aspects underpin overall comprehension and show where RSVP falls short. Where RSVP shines is for short bursts of text (as the article states) and also for learning to read words in groups rather than individually, which is really really useful.

But back to my original point, speed reading isn’t just RSVP. It’s a collection of techniques that together help you read more effectively. Honestly, it probably shouldn’t be called “speed reading” at all - it’s what naturally strong readers already do. You solve for speed, comprehension and retention.

You start with the mechanics (learning to read words in groups, strengthening eye muscles, improving concentration etc) so that you aren’t getting hung up on bad habits, and then assess how well you understand and remember content. In fact, when you learn to read in this way, you don’t even need the tools like RSVP and moving highlighters. Yes, they are useful training wheels, but you’ll already read that way anyway. You also won’t be tied to one service or medium.

Believe me, I find the hyped-up, late-night infomercial, eat-cake-get-thin approach really off-putting too, in fact it’s one of the reasons I made Acceleread in the first place, but don’t throw the baby out with the bath water.

[iandanforth]

Using a "let me google that for you" link on the problem with software patents seems unnecessary. Why imply your users are lazy? At first I thought the contrast between "whose patents are pending" and that link was a great bit of commentary, but that link was an unexpected slap. I was expecting an EFF link.

[pkghost]

Oops—did not wake up wanting to slap you :) My thought was more along the lines of "if you're not aware that this is a problem, here's a list of big/obvious headlines to pick from, or just walk away with the sense that there is an active conversation about the issue"

[Touche]

Really like it. The only downside is that you have no idea how far along in the article you are. Would be nice if there was a % complete shown somewhere.

[mallamanis]

I'd really like to see this model with an adaptive speed. Same speed for all the words isn't always the best.

Maybe using an n-gram model to predict how probable the next word is, could pass high probability phrases at a faster rate, while slowing down for "harder" words

[lsb]

There's a good Google Books dataset for doing exactly that, and you can doing something in Hive to get you the lowest-entropy n-grams (with something like https://github.com/lsb/text-entropy/blob/master/passphrase-s... ), and then you can the low-entropy n-grams into a Bloom filter (like http://www.leebutterman.com/passphrase-safety/how-it-works.h... ) and your enormous corpus gets fitted into a few dozen megs of memory.

(That sounds like a cool idea, hit me up if you're game for hacking on something like that)

[johnwatson11218]

I always plug the same site on these threads. www.zapreader.com/reader I use that several times a week to get through lengthy articles. What I really want is something that shows all the text in the background kind of blurred out. The words should still flash on screen but with a keyboard shortcut I can make that disappear and have the word I was on hi lighted. I would like the text in the background to scroll while it is blurred out and I'm speed reading. That way if there is an image or diagram I can quickly shift back to normal reading. Or if I just need to re-read a section. Then using only keystrokes have it resume blasting the words on screen.

[wpears]

Similar to what I did with the chrome extension Spree [1]. Though I find your pauses on periods to be a bit much. Code for Spree available on github [2]. The IIFE in spree.js can also function as a bookmarklet.

Also, Spree doesn't walk to an element's parent, which usually keeps it from getting into JS and ads, while still reading all of, say, a news article.

On another tack, quite a lovely site.

1. https://chrome.google.com/webstore/detail/spree/aehoaolhojlm...

2. https://github.com/wpears/spree

[myrdev]

I really like the way this works - much less disruptive to jump in and out.

Any reason you cap WPM at 666, and always default to 400wpm? I want to go fasterrrr :)

edit: I made my branch do this, are you interested in contributions?

[ScottWhigham]

Great stuff - thanks for sharing.

One note: when I drag the bookmarklet from the Install page to my toolbar, then try to use it on a page on my http://localhost/ (a dev site, for ex.), it (a) does what I expected it to do, but then (b) forwards me to http://localhost/install.html when it is finished. When I browse an actual domain-based site, it gives me the nice "You just read..." message at the end. It would be nice if both local and remote sites had the same experience.

[pkghost]

Welcome to my smelly code! It's supposed to go to /install on squirt.io--and also localhost for my own dev purposes. It's a bit of a hack, but saves me from maintaining two bookmarklets. Are you a publisher? Trying to embed it on your blog/site?

[coderzach]

I feel like it ruins the cadence, in my head anything I read with this sounds like it's coming from a robot.

[javanix]

Yeah, I tried reading an excerpt from http://deadspin.com/when-englands-greatest-genius-took-over-... and not only did it sound odd in my head, I felt like I wasn't really reading any quicker than I normally would.

I think that perhaps a line-by-line reader would be appropriate, rather than word per word.

[wpears]

You're quite correct that cadence is diminished, though that can be mitigated by adding a delay to each word based on its length.

[gchokov]

I'd be much more interested in a solution that effectively minimizes the text or uses another technique about speed reading (like diagonals, etc).

The reason why I won't be using this service is simply because it makes the eye lazier. My eyes are already lazy enough because I am in front of the computer 12+ hours a day, so my eyes muscles need movement. Staring at one point for long time can also cause side effects like losing the sense of space and time. Staring at one point is often used as hypnotizing during different kind of therapies.

[felipeerias]

I am not sure that I get the point of this. When I read a long piece, I try to enjoy it. Imagine the same approach being used to "fix" food or sex in the name of efficiency.

[talkingquickly]

I think the use case for this is the reading we do because we need to consume the information as quickly as possible rather than the reading we're doing for pleasure. Not that the two are mutually exclusive by any means.

I love reading but, particularly online, I'll often need to read something which contains a lot of information I need but isn't written in a style I particularly enjoy reading. For these, I can see the benefit of tools like this.

[Jonovono]

Just because you are reading quicker does not mean you are not still enjoying what you are reading..

[arikrak]

Have you heard of Soylent?

[dgfvd]

There's a bug that I can't really fathom:

when reading this paragraph:

"The WEIRD mind also appears to be unique in terms of how it comes to understand and interact with the natural world. Studies show that Western urban children grow up so closed off in man-made environments that their brains never form a deep or complex connection to the natural world. While studying children from the U.S., researchers have suggested a developmental timeline for what is called “folkbiological reasoning.” These studies posit that it is not until children are around 7 years old that they stop projecting human qualities onto animals and begin to understand that humans are one animal among many. Compared to Yucatec Maya communities in Mexico, however, Western urban children appear to be developmentally delayed in this regard. Children who grow up constantly interacting with the natural world are much less likely to anthropomorphize other living things into late childhood."

it seems to stop displaying words directly after the U.S., string. I had to restart it after that.

Another note: words that contain hyphens or em dashes are not tokenized so you get huge words that are impossible to read at speed.

[bra1n]

While the idea is nice, I'd rather not use a bookmarklet that tracks me through a unique ID across websites and not only keeps a list of pages that I used it on, but also of pages that I came from: https://github.com/cameron/squirt/blob/gh-pages/bm/squirt.js...

[AlexSolution]

I've tried several of these, and I really think jetzt https://github.com/ds300/jetzt is the best one. It has a progress bar, keyboard shortcuts, and I particularly like the way it wraps words in enclosing elements - like quotations and parentheses - when the word being displayed is within the enclosure.