Fixing the Web One Page at a Time, or Actually Implementing xkcd #37

Thomas Steiner, Ruben Verborgh, Rik Van de Walle

Spot the errors

"“The World Wide Web - a system of interlinked hypertext documents, a provider of endless knowledge and a source of kitten pictures - is unfortunately seriously broken..."” :-(  

Common pollutants on the Web

• Dots instead of ellipsis
• Missing Oxford comma
• Hyphens instead of dashes
• Typewriter quotes
• Bad line-breaking

…and most importantly:
the [adjective]-ass [noun] case

This
is a serious situation.

What if the Web could fix itself one page at a time?

There is
an app
for that!

Introducing the xkcd#37 extension

• Chrome extension at bit.ly/xkcd37
• Highly customizable
  • deals with all of the aforementioned pollutants
  • plus your personal annoyances
• Enjoy the Web again
• Browse without fear
• Fixes the Web once and for all

Easy to fix with regular expressions

• Example: ellipsis
  • \.\.\. becomes …
• Example: avoid orphan ‘a’s
  • (\ba)\s(\w+\b) becomes $1 $2

How to fix that dumb-ass xkcd #37?

• Regular expression with function replacement
  • [adjective]-ass [noun]
  • (\w+)-(ass)(\s+)(\w+)
• Part-of-speech tagging (Brill, 1992)
  • $1 should be an adjective
  • $4 should be a noun
  • donkey-ass picture: OK
  • smart-ass picture: smart ass-picture

How to get all the text nodes?

• XPath
  • the classical option
• Recursive
  • the hand-crafted option
• TreeWalker
  • the cool API option

Let's test which one is the fastest: bit.ly/xkcdperf

We still need to fix dynamic pages

• Static pages have been covered
• Dynamic pages are omnipresent
  • Twitter, Gmail, and many others
• Listen to DOM events
  • DOMCharacterDataModified
  • DOMSubtreeModified
• Special attention to <title>

Enough. Let's fix the Web!

• Install the extension at bit.ly/xkcd37

Isn't that awesome?

• Let's check Twitter

Tired of browsing a broken Web?

• Change your life with bit.ly/xkcd37
• Share your annoyances on Twitter #xkcd37
  • @tomayac
  • @RubenVerborgh
• Fixing the Web? Yes we can!