Quora looks to the crowd to solve machine learning problems

There’s a lot of question-and-answer websites for internet users to check out — from technically oriented options like Stack Overflow to very broad platforms like Yahoo (s yhoo) Answers. Quora wants to stand apart, not necessarily by attracting the most people to it, but by being smart about what it surfaces.
So the company is hosting a machine-learning contest next week for developers to take real production data from Quora and cook up new models that could help improve Quora’s service. The upcoming eight-hour online event is being run on HackerRank, a site for coding contests.
Quora has done this type of event before. One previous contest focused on doing a better job of separating the good answers from bad ones. A developer in India working under the username kahtras came out on top, although his submissions are not public. This time around, Quora engineering manager Kah Seng Tay wrote, “[W]e’re now ready to offer a deeper set of real-world problems with data from Quora’s production databases.”
Quora isn’t unique in looking to the crowd for new solutions. Facebook, (s fb) Ford, (s f) GE (s ge) and other companies have run contests for data-science challenges on Kaggle, while NASA and other government agencies, as well as the Harvard Business School, have taken the crowdsource route on Topcoder. And then there was the Netflix Prize (s nflx) to develop the best possible recommendation engine and the company’s more recent Cloud Computing Challenge.
In these sorts of competitions, real production data is often made anonymous — Quora did this for its good-and-bad-answer contest — although researchers have managed to undo that process in the past.
While machine learning is edgy, it’s not as if the skill set is limited to a small group of people. People can familiarize themselves with the basics of machine learning and other hot topics on Coursera or Udacity and get a shot at catching attention from Quora employees, and perhaps even a job offer, if they’re really good.
Feature image courtesy of Shutterstock user Carlos Castilla.