Riiid (CEO Youngjun Jang), an EdTech Startup, announced on the 16th that the company applied for a patent on a question recommendation algorithm that provides test takers with customized multiple questions they need to review by using the machine learning technology.
Solely using the machine learning technology without human intuition, this patent predicts test takers' answers, right or wrong, analyzes the result data, and calculates the individual’s understanding on its question set. The technology prioritizes and sets questions that a test taker is not familiar with, or questions that a test taker have not understood before, to lead them to achieve their target test score.
Its prediction accuracy rate has proved to be very high, showing more than 90% on questions that the users that gave the right answers. The key of the patent is that the algorithm identifies and determines necessary questions and unnecessary questions for each individuals to review, after completing a customized user data analysis. This means that it can prioritize and provide questions in the order of necessity and urgency, by putting low priority to the questions with high probability of giving right answers, but high priority to the questions to needed to study.
CTO Jaewe Heo said, “Our patent goes in a direction to replace human intuition with machine learning at a maximum level. Going forward, we will apply for additional patents on more advanced new algorithms based on current test question recommendation solution. The key is how to upgrade and enhance the algorithm. As Riiid accumulated more than 15 million case data and owns its machine learning source technology, next patent applications will be possible in the near future.”
CEO Youngjun Jang explained, “No other EdTech company has this large amount of question-result analysis data as Riiid does. We are co-developing the algorithm with Professor Changho Seo, at KAIST, who is actively conducting many studies in ICT and machine learning area.
Original Article: http://www.fnnews.com/news/201611160908266595