Quality Project - Post characteristics as images?

Question

In the CoGro Musings chat room, Tim Post mentioned

In the quality project, we started examining the 'shape' of a question by turning it into a picture that its characteristics painted. Vowels become green blocks, consonants blue, punctuation purple, space black, code orange - you get the point. Running this against posts with a known good / needs help history showed it could actually prove to be deterministic. Someone with a bit of time might enjoy picking that up.

This sounds very interesting.

Was the route of turning it into a picture taken due to the large number of features you were looking at? How do these pictures compare to other output types in terms of usability? I'm used to seeing our data turned into scatter plots, bar charts or line plots, so that is my comparison point.

Would it be possible to share some of these images (mostly for my curiosity)?

Answer 1

I like using regexes. I wrote one that, when used on regex101.com, makes pretty pictures like the ones below. It's a different color theme, but it helps show the idea off. It's also fairly simple, and there are a few edge cases that aren't handled, but it works well enough in most cases.

You need to have the markdown for code to be detected. You also need to use the gmi modifiers. But you can easily test on your own this way. I haven't had the opportunity to analyze much, but I'd be interested in any patterns you find.

Regex with named groups:

(?<code>    .*|`[^`]+`)|(?<vowels>[aeiou]+)|(?<punctuation>[[:punct:]]+)|(?<consonants>[bcdfghjklmnp-tv-z]+)|(?<spaces>\s)|(?<digits>\d+)

Same regex without named groups:

(    .*|`[^`]+`)|([aeiou]+)|([[:punct:]]+)|([bcdfghjklmnp-tv-z]+)|(\s)|(\d+)

Pictures:

Your question:

One of my SO answers _{^{(this answer only has +5 score, which is the highest I've gotten on SO... Consider voting for my answers on SO if you like my regex here)}}:

A random question scoring -5 on Stack Overflow: