iTalk2Learn: The Next Generation Virtual Tutor

Whizz Education is committed to leading research in innovations to enhance the Maths-Whizz Tutor and, ultimately, your child’s learning experience. That’s why we’re part of iTalk2Learn: a 3-year EU-funded research project with the ambitious goal of creating a next generation virtual tutor.

The iTalk2Learn tutor seeks to replicate the behaviour of a human tutor. Using cutting-edge technologies, like speech recognition and emotion detection, the virtual tutor will be able to identify and respond to your child’s learning needs more closely than ever before. (It sounds like something out of a futuristic sci-fi thriller.)

As a commercial partner, we are ideally positioned to test ideas and exploit the results of the project. We are currently actively involved with two key components of the iTalk2Learn tutor: speech recognition and the adaptive sequencer.

Speech Recognition

Over the past few months, we’ve collected several hours of voice recordings from Maths-Whizz students to help train the speech recognition model. By gathering and analysing these recordings, our partners will work to create a model that detects how students are getting on with their lessons using audio cues. For example, if a student says they are ‘bored’ or finding the content ‘too hard’ during a lesson, the model will detect and interpret these cues, then instruct the tutor to move them on to a more appropriate lesson.

This experience has opened our eyes to new possibilities for student-computer interactions. At the same time, it has highlighted some potential limitations of speech recognition, one of which is that students typically say very little to a computer. We must therefore also aim to engage students in dialogue so that the tutor can pick up audio cues in the first place.

Maths-Whizz Exercise

The current Maths-Whizz Tutor mainly communicates with students using text inputs. Through iTalk2Learn, we’re looking at ways to engage students through spoken conversations.

Adaptive sequencer

Have you ever wondered how the Maths-Whizz Tutor works out which lesson your child should do next? The answer lies in our adaptive sequencer, which takes into account your child’s unique performance history to find the lesson that is most appropriate for them. The idea is to keep your child in their so-called Zone of Proximal Development (in other words, the ideal learning state, in which they are challenged just the right amount).

iTalk2Learn’s Vygotsky sequencer aims to take the Maths-Whizz sequencer to the next level. It is rooted in sophisticated algorithms that use more data to determine the next lesson for your child even more accurately. This approach is similar to the way that Netflix recommends films to its customers. Netflix looks at the films a customer has previously watched, as well as those that other customers with similar taste have watched, and recommends new films based upon this analysis of their preferences. The more movies you watch, the more accurate the recommendations become. In other words, as you interact with the site, Netflix learns about your interests and adapts to them.

In the same way, the Vygotsky sequencer studies each child’s lesson history, as well as the lesson history of students with similar profiles, and recommends the lesson that is ideal for that child. The more lessons completed, the more accurate the sequencer becomes. Win-win.

In June, we ran a month-long trial with Rosemead Preparatory School to test the Vygotsky sequencer. By randomly assigning the sequencer to about half of Rosemead’s Year 3 and 4 students, we were able to compare the Vygotsky sequencer with the current version in the Maths-Whizz Tutor. We have already learnt that, while the new sequencer may have delivered lessons at the right level of difficulty, this might mean that it compromised the students’ experiences by repeating some topics more than the current Maths-Whizz Tutor.

Putting customers at the centre

Both the voice recordings and the Rosemead sequencer trial highlight how important it is that our customers’ real experiences are what drives the direction of this research. The science of speech recognition and machine learning is elegant, but its power is limited by the practical day-to-day realities of our students.

Throughout the iTalk2Learn project, and in all of Whizz’s research efforts, we put students, parents and teachers at the centre of everything we do. It not only makes commercial sense but, more importantly, it is the only way to deliver the best possible educational experience for you, our customers, and that’s exactly what drives us at Whizz Education.
Whizz Education and iTak2Learn logos

Visit to learn more about the iTalk2Learn project. If you have any questions about Whizz Education’s involvement in the project, please contact Junaid Mubeen at [email protected].

Find out more about virtual tutoring

Subscribe to our parent newsletter!