Sramana Mitra: What’s happening on the technology end? How are these questions generated? How is the progression being determined?
Matthew Glotzbach: Most of the interactions are with texts and images. We do a lot of natural language processing and understanding of the text. I’ll take a multiple choice question as a simple example.
Let’s use a simpler domain than organic chemistry to illustrate the point. If you’re trying to learn the capitals of the 50 states in the US or the capital cities of the countries around the world, we’ll say, “What’s the capital of California?”
We’ll generate smart distractors. You probably know that New York is not the capital of California, but you might think that Los Angeles or San Francisco is. We’ll generate those distractors as they’re called. We’re using technology to generate those questions and generate different types of questions. Obviously, that’s a straightforward example.
We’re also leveraging these billions of data points that we get on a regular basis. We’ve asked hundreds of thousands of other people this month what the capital of California is. We can see what people are getting right and what people are getting wrong. We can get an intuitive understanding from the data. We can use that to train our algorithms to know how to better sequence things and adjust the level of difficulty.
Those are the two things that are happening. A lot of natural language processing and intelligence is used to generate questions. You can imagine a fill-in-the-blank question where we’ll give you a paragraph passage about a historical event. We blank out the key concepts because we can understand the context of that. Then, we use the data from all of the study that happens on our platform to train the models to better personalize and adapt.
Sramana Mitra: Is everything machine-generated at this point?
Matthew Glotzbach: Yes and no. We start with the content that’s provided by users. All of the initial content is really from the user. Then we’re asking the machine to do the work behind the scenes to take that content and do interesting and useful things with it.
Sramana Mitra: You don’t have content experts or subject experts developing content?
Matthew Glotzbach: That’s correct. We’re a supplementary learning platform that people use in conjunction with whatever course that they’re taking now. That can be a traditional classroom-based course, but it can also be an online course.
You might be using a popular language learning app like Duolingo. You might use Quizlet in conjunction with that app to make sure that you are learning all the vocabulary you need to learn.
Sramana Mitra: How do you charge?
Matthew Glotzbach: It’s a freemium business model. The vast majority of our users use the platform for free. We have some light advertising. Then we have a premium subscription upgrade. They can buy to get access to some advanced features including removing ads.
Sramana Mitra: How much is that subscription?
Matthew Glotzbach: We have two student subscriptions. The basic subscription removes advertising and makes the mobile apps work offline. It’s $12 a year. Quizlet Plus layers on advanced analytics. That’s $2 a month, so $24 a year.
This segment is part 2 in the series : Thought Leaders in Online Education: Quizlet CEO Matthew Glotzbach
1 2 3 4 5