“I hope it will help others”

“I hope it will help others”

Visually impaired student transforms “far too much” information into comprehensible podcast

05-02-2026 · Background

Long reading lists and longer glossaries. For many students, it is a challenge to learn it all by heart, but even more so for Noortje van Maldegem. The second-year Data Sciences and Artificial Intelligence student, who only has 20 per cent vision, was looking for – and found – a simple but effective way to get to grips with the material.

“Podcasts!” Van Maldegem reveals the secret immediately. She makes them herself, although she says that does sound more significant than it is. Her podcasts are generated by a computer and are able to break down the course material thanks to information entered by the student. “It’s not very exciting, but it is very effective. And an absolute game-changer for me.”

It all started with her Computational and Cognitive Neuroscience class, says Van Maldegem. “A fantastic subject, but there is a lot of literature. I would spend a whole weekend working through the reading material for just one lecture.” The topics discussed turned out to be complex, full of terminology that Van Maldegem didn’t know. It also takes her longer to read through everything as it is. Her impaired vision is the result of a tumour on her optic chiasm, where the optic nerves intersect.

Walking

She asked her lecturers whether there was an alternative that might allow her to keep up, but they didn’t have any ready solutions. Then something gave Van Maldegem pause for thought. “I walk to the university and back every day, about a six-kilometre round trip. What if I were able to listen to the course material while I was walking?”

Initially, the student tried university software which converts text into speech and is offered to students who might benefit from that. “But it wasn’t that great. I wasn’t even able to upload all the material I wanted to convert. And the speech it was able to produce was very cold and synthetic. Not something that will help you remember.”

Conversation

More natural speech, as if somebody were talking to you and explaining things, would have a much greater impact. “Like on a podcast.” This led Van Maldegem to NotebookLM, Google’s AI-assisted note-taking tool which analyses and summarises documents, such as PDF files. She selected the most important pages of the book, added the lecture slides and the glossary, and generated the first podcast in Dutch – a promising result. “It was just like a conversation between two people talking about the reading material.” Van Maldegem showed the result to her teacher Michael Capalbo, a member of the Faculty of Psychology and Neurosciences, who applauded her initiative and encouraged her to write a report about her findings, which would then count as an extra assignment for the subject.

Van Maldegem discovered that the podcasts were even better when generated in English. “In Dutch, they were quite superficial and usually lasted less than ten minutes. In English, they varied between 21 and 48 minutes, and were far more in depth. Not that strange, when you consider that many of the AI models are trained with English data.”

Value

Now the next step is to broaden the accessibility of the application. The idea is to present the concept – together with Capalbo – to Disability Support (which supports students who have physical, mental or sensory disabilities, or a chronic disease or condition). “We haven’t got that far yet, though,” says Van Maldegem, who is particularly keen to help other students. The budding data scientist is unsure whether she wants to develop an independent tool (unrelated to Google) to create podcasts. “I’m not that proficient yet. There are also so many already, and developments are so quick, that I don’t know what value I can add to the field.”

Photo: Observant/Illustration: Shutterstock

Tags: noortje van maldegem, podcasts, visually impaired, disability support, AI, Google, tool, data science,instagram

Add Response

Click here for our privacy statement.

Since January 2022, Observant only publishes comments of people whose name is known to the editors.