X-Europe Data Science Webinars

Modeling difficulty of English exercises - probabilistic programming approach

Damian is an Associate Data Scientist at Pearson IOKI Poznan office,
where everyone are passionate about and committed to
creating e-learning platforms and English learning and teaching materials.

Selecting test exercises with an appropriate difficulty level is one of the most important aspects of a high-quality assessment, but how do we know how difficult a given exercise is?

One of the possible approaches is to use Item Response Theory - a set of psychometric models which
can be used to estimate learners’ proficiency and parameters of each exercise. This approach works well in most cases, however, is data intensive. So, the question arises what to do when we don't have
enough data, when for example, we are introducing some new exercises.

Fortunately, probabilistic programming comes to our aid. Using Greta, we can build a custom IRT model, that allows us to quantify how different features of exercises impact their difficulty. In this talk, we will
introduce the audience to the basics of IRT and probabilistic programming through a real-world application - estimating the difficulty of English tasks. For this purpose, we will use the previously mentioned Greta package, which lets us write statistical models in native R and sample from them automatically and efficiently.


And Barcelona Data Science and Machine Learning Meetup,
Budapest Deep Learning Reading Seminar,
Budapest Data Science Meetup