Could this Algorithm Guess the Name of your Future Baby?
"The age of baby name books is over", says Roy Golan, developer and co-founder of The Namestork, an app aiming at speeding up baby naming process
Collaborative filtering is a rising algorithm based search technology used nowadays in countless services. Whether it's books, movies or fashion items you're looking for, collaborative filtering allows finding more items to match your taste, and it does so much faster rather than the traditional browsing methods. The implementation of this approach in baby names apps was just a matter of time, and now an app called The Namestork does exactly that.
Roy Golan, one of the app's developers, says it was created to change the way we look for optional names. "We saw many cases where parents have a 'feeling' of the name they're looking for, but still can't find the one to really match their new born." This was the beginning of the baby names recommendation engine. "Naming a baby is a process", says Golan. "It usually begins with creating a list of options, and just to get to this list people are scanning baby name books and websites for hours. 95% of the names listed on these platforms are totally irrelevant to the parents' specific taste. We hope to change that with The Namestork, and our user-driven data shows that it's already happening".
Can an algorithm really understand human taste?
The Namestork's engine was created using a list of the top 30,000 first names world-wide and years of user name preference data. Whenever two names appeared on the same list, that info was added to the matrix, and the cumulative result allows generating the most similar names for any user input. After getting the initial recommendations list, the user can rate names and fine-tune the next 10 results. Eventually, the algorithm gets an established profile of the future-parent's taste, so that all of the relevant names are on the screen in no time.
"The algorithm is getting smarter with every use", explains Golan. Parents from modern secular societies will often look for names that are somewhat unique, and that makes exact prediction much harder. To overcome this challenge, the recommendation algorithm is designed to keep learning about name preferences all the time. "The app’s algorithm shows some surprising abilities, given that it is purely statistical and mathematical, and doesn’t really ‘understands’ names sound, meaning or origin. Entering the names Byonce and Jayla, for example, will produce the names Meyonka, Neida, Heini, Alys and Ayanna. Typing in Wolfgang and Friedrich will generate Helmut, Karl-Heinz, Lothar, Bernhard and Siegfried".
In addition to the recommendation engine, the app provides a convenient baby name list management tool, making it easy for users to gather names with one click, and then share the final list with friends and family in multiple formats. Another section of the app contains a comprehensive list of typical names around the world, consisting of the top 100 typical names in each and every country.