News for hackers. Stuff that matters.
Re:30% of Google's Emotions Dataset Is Mislabeled by LanceH on Thursday July 14, 2022 @10:21AM (#32096410)
In my first few months as programmer in the 90's, I realized that human inputs were sketchy. It was from a form field for which state someone was from (USA). Fifty possible states, 2500 different entries. Sure there was a bit of garbage, but 95+% were recognizable states...but why random capital letters? How do you get a space in Wyoming?

It was a good lesson at the start of my career that I see playing out over and over. When I see some cool demo of statistics across the country or globe, I'm more impressed by the effort of cleaning the data than the stats behind it.

There are no comments on this comment yet.