The tag clouds on Richmond Sunlight have been having a little trouble. Given that a quarter of all General Assembly bills are commendations, the word “commendation” has been so enormous as to dwarf everything else. So I’ve had to artificially cap the size of that tag to prevent it from taking over. Next time your legislator complains about the sheer number of bills that she faces each year, ask her how many commendations she introduced this year.
Comments are closed.
Why cap it then?
I went back and forth on this for a bit. I decided to do so because all of the font sizes are relative. If we have ten tags that represent between 5-10 bills, then the tags that represent 5 bills might be in a 10 point font, the tags that represent 8 bills might be in a 16 point font, and the tags that represent 10 bills might be in a 20 point font. So if there’s one tag that represents, say, 500 bills, and all the rest represent between 1-50, there’s one word that’s enormous and a whole bunch others that are so tiny that it’s difficult to discern the difference in their sizes. Without “commendation,” the difference between the largest and smallest bills might be a 200% font size difference. With it, ignoring the “commendation” tag, there’s a difference of, say, 10%.
So it comes down to a simply problem of usability. Tags just aren’t real useful if there’s a whole bunch of teeny-tiny words and one huge one.
Waldo, you might consider using a log() or sqrt() function on the font sizes. I’ve seen that used to great effect on a couple sites.
(Square root makes sense because it keeps the word’s area proportional to the tag’s popularity.)
That’s too funny.
Did you tag them all yourself, Waldo?
I was going to suggest something along the lines of what Tim McCormack says. Ironically, just yesterday our development group was facing the same issue. We have a tag cloud with 10 font increments but one particular tag dwarfs the rest. It has ~120 occurrences which has the net effect of requiring every increment to be 10%, or 12 steps. Therefore an item tagged once or 11 times has the same font.
Now that I’ve described the problem, we believe the solution is to implement a sloping scale where the “step” is relative to the mean point. I haven’t worked out the math for it, but we are going to take up the issue soon. Maybe we can work together on it? Our application is Ruby based.
Many of them, but an increasingly large majority were tagged by site users. I even saw one legislator tagging his own bills yesterday, which is pretty great.
Duane and Tim, I’m real interested in your ideas about proportional sizes. Using a logarithmic scale makes a lot of sense. I’ll play with that after work this evening.
This logarithmic scale is much better. Thanks so much for that suggestion!
For the record, the logarithmic scale (predictably, had I thought it through) works best in those tag clouds that have an enormous disparity between the largest and smallest tags. For those that don’t have, say, twenty-fold gaps, a sqrt scale is far more useful.