The first two accusations are reasonable – there are a lot of stupid comments out there, and they are very repetitive.  They are so hated by some that someone has developed a Firefox Add-on that will strip comments out of the page before you can even be offended by them.  On the other hand, though, people are very stupid and repetitive.  There’s often a surprising honesty to the statements that people make on youtube. It is suggested that the monumental scale of the youtube ‘community’ means that individuals are effectively anonymous, which liberates users from any self-censorship that would occur from fear of shame or other punishment.  It’s not all honesty, since people are more likely to be offensive if they know they can get away with it, but it’s a refreshingly different place to look for views and opinions in.

So while they may be stupid, they’re still worth looking at.  The issue of quantity is more of a practical issue, though.  For example, I keep tabs on  a video called “The Most Pathetic Baby Panda Ever”. It has almost 5 million views and 15,500 comments.   Even if you only dedicated 5 seconds to each comment, you have to spend over 21 hours studying them. The Evolution Of Dance has almost 250,000 comments -   almost two weeks’ worth of study.  While they’re interesting, they’re not that interesting.

Conveniently, though, this is exactly the kind of thing that the study of data mining is supposed to deal with.  In addition, the fact that the comments appear to be stupid (or simple,) and highly repetitive works in our favour. The first question I would try to ask of these comments is “what’s a typical comment?”  and toward that end, I have created the Youtuber.

Most of these factoids are interesting only to myself, sometimes not even that. Being able to dig into my financial transactions gives me the opportunity to do a number of things, though – I can construct an “average” day. If I place my expenditures on a city map, I can draw my route over a day (or a week, or a month), see when I buy my morning coffee, and the average radius of my lunchtime wanderings. I can even use it to retrace my steps, and find the name of that great Japanese restaurant I went to in Auckland last year.

You might still think that these are unimportant, but this ignores two issues. First, we don’t know what’s insignificant until we can see it. There may be important trends in my buying habits – if I buy a coffee in the morning, I may end up working late because I forget to go home. Second, it’s important to somebody – it might not be front-page news if I go to a restaurant, but it’s nice to have the name on hand so I can tell a friend whether to give it a miss or not.

Most of our interaction with computers (as digital cameras , cellphones or point-of-sale devices) is being recorded. The sum total of this recorded data is referred to as our digital footprint. While many people find the presence – or even the idea - of this record threatening, I think it gives us an opportunity to answer important questions about who we are, and what we do.

I was trying to make a function that would be able to indicate the correlation of two keywords to one another over the volume of youtube comment posts.  That is, if 50% of comments contained the word “awesome” and 50% of the comments contained the word “lol”, then it would be expected for about 25% of the posts to contain both.  I decided that green should indicate “about expected” correlation, cyan “above average”, blue “very high”. On the other side, I would choose yellow for “below average”, and red for “very low.” 

Messing about in paint, I figured out where I wanted my red, green and blue values to move in and out, and once I had those states working I would interpolate between them.

just another gradient

The colours didn’t work out properly at first, but it I like the gradient here anyway. It’s interesting when the mistakes look as nice as the successes.

This is the state of the links in the youtuber as of today. the colours here are for values between 0.25 and 6… because the function is piecewise, nothing interesting happens past 2.0… it was gratifying to get it working though.