Most likely, you’ve been told that your name does or doesn’t suit you, but what does that even mean? What does a Rachel look like to you? Is there anything to the idea that certain names naturally go with certain faces?
Perhaps, say some researchers at Cornell. They note that while names seem random, they’re not. A person’s name depends on their gender, their ethnicity, their culture and the popularity of certain names at the time they’re born. In 2011, the most popular baby name for boys in the United States was Jacob and for girls was Sophia. In 1880, it was John and Mary. Minnie was number five on that list for girls in 1880, but it hasn’t been in the top 1000 names in the past thirteen years.
So what the scientists at Cornell decided to do was build a database full of faces with names and train a computer to pick out the commonalities. They picked out one hundred faces from Flickr, using names that represent something like 20 percent of of people in the United States. Then they showed those faces to people and the computer algorithm to see who was better.
What they found was that the computer could predict the name of the person 4 percent of the time. Which isn’t a lot, but is four times the accuracy of a random guess. Here’s New Scientist on what this sort of technology could do if it worked better:
With improvements in accuracy, the software could be used to automatically harvest the names from online picture captions to tag the people in the picture, Gallagher suggests. The work will be presented at the Computer Vision and Pattern Recognition conference in Portland, Oregon, in June. Gallagher says he intends to expand the database to encompass all known names in the US.
New Scientist lets you try to guess the names of their staffers here. Are you better than the computer?
More from Smithsonian.com: