my eyes got killed when I click on the link. bleeding orange. but got used to after a minute.
how do i check other users?
some minor glitch in html/css - on chrome top banner glues text together (I gave adblock turned on fyi)
last q: how do you scrap data??
Yeah design should have been better done. To check the other users, hover over the username for them. Should have been clearer and tested better overall.
What is the "average comment tree"? If one has a forest (a disjoint union of trees), how does one 'average' them? What do the tooltips mean? It's not clear to me what I'm looking at here.
I think this visualization has potential (I have similar ideas along these lines) but at the moment it feels a bit like "visualization salad".
Agreed on the visualization salad comment. This isn't a normal bar chart and there are some details that would be nice to know.
Such as: comments on hacker news are trees with the added property of having order within each level. So you can "add" trees together by overlapping the trees and calculating how many times a node occurs. Naturally, the top nodes will have always have higher percentages than the lower nodes within each level. The interesting part of it to me was comparing the trees produced for each user and the number and types of comments of people submitted.
The tooltips display information about the types of comments that occured at that position. The first piece of info is who commented most at that position for that particular user. So the first node is always by the main user. If you look at pg's tree, then you see that the top comments are basically pg talking to himself. We then see the chance that the node will occur at all. So with pg, there is a 64% chance that someone will reply to him at all. Then after that, there is a random comment displayed at the position to see the types of things might say at that point in the comment structure.
If I had to do it over, I would use the reddit API instead of crawling hacker news and regexping the data out of the html. I felt pretty guilty doing it and the data wasn't all that great. Also, I was testing this mostly in chrome and just noticed that the tooltips break in an awful way with firefox. Poop.
6 comments
[ 2.5 ms ] story [ 23.2 ms ] threadAnd I crawled hacker news with python. It was shameful and I am not proud of it: https://github.com/zmaril/HN-Visual-Comments/blob/master/web...
I think this visualization has potential (I have similar ideas along these lines) but at the moment it feels a bit like "visualization salad".
Such as: comments on hacker news are trees with the added property of having order within each level. So you can "add" trees together by overlapping the trees and calculating how many times a node occurs. Naturally, the top nodes will have always have higher percentages than the lower nodes within each level. The interesting part of it to me was comparing the trees produced for each user and the number and types of comments of people submitted.
The tooltips display information about the types of comments that occured at that position. The first piece of info is who commented most at that position for that particular user. So the first node is always by the main user. If you look at pg's tree, then you see that the top comments are basically pg talking to himself. We then see the chance that the node will occur at all. So with pg, there is a 64% chance that someone will reply to him at all. Then after that, there is a random comment displayed at the position to see the types of things might say at that point in the comment structure.
If I had to do it over, I would use the reddit API instead of crawling hacker news and regexping the data out of the html. I felt pretty guilty doing it and the data wasn't all that great. Also, I was testing this mostly in chrome and just noticed that the tooltips break in an awful way with firefox. Poop.