Approach
Zeeguu helps learners pick up vocabulary in three ways:
- Finding articles on the web that match each learner's interests and level
- Supporting reading and listening with in-context translation and pronunciation
- Reinforcing the words a learner looked up with personalized exercises and audio lessons
Zeeguu runs on the web, on iOS and Android, and as a browser extension. All the code is on GitHub.
Zeeguu in numbers
All-time totals as of September 2026.
Monthly active users over the last year, live from our API:
Usage follows the academic year: classroom use drops over the summer. Most learning time is spent reading; see the activity breakdown.
Papers
Linked PDFs are the authors' versions; cite the published version where there is one.
A catalogue of the patterns that show up when a slow, costly, and non-deterministic LLM becomes part of an interactive app that has to be fast and trustworthy. The patterns come from integrating LLMs into Zeeguu.
Learners who read with one-click translation show the words they don't know by looking them up. Over time they also show the words they do know: a word met again and again across articles without ever being looked up is evidence that the learner already knows it. The paper validates this signal and shows how per-word estimates add up to a vocabulary-growth curve.
How to estimate reading time when studying free reading in the wild.
Uses Zeeguu to teach vocabulary during moments of online procrastination.
Introduces the project and reports on its use in a high-school French class in the Netherlands. Students liked it and made use of the chance to read about topics they cared about.
The general ecosystem architecture that Zeeguu is an instance of.
Talks
Student theses
In a controlled experiment with 22 participants, users found audio-focused exercises more fun and said they were more likely to keep using Zeeguu.
An architecture for the Zeeguu browser extension. Half of 32 participants preferred the extension's cleaned-up article view to the original page. Also discusses what it takes to maintain such an extension.
Milestones
- 2026
- The iOS app launches in January and the Android app in February. A fresh audio lesson built from each learner's own words is now ready every day. Teachers can run a class that sees only the texts they share. Learners can choose a language variety (Belgian or Netherlands Dutch, Brazilian or European Portuguese), and Bulgarian becomes available. Word lookups pass half a million.
- 2025
- Romanian, Portuguese and Greek become available. LLMs enter the platform: personalized audio lessons built from the words a learner has looked up (July, for everyone from September), simplification of any article to the learner's level (August), and topic classification. Learners complete their 400,000th exercise.
- 2024
- A multilingual topic classifier goes live, and learners start getting recommendations based on articles they liked before. More than 2,000 learners have now used Zeeguu.
- 2023
- A Villum Experiment grant funds infrastructure work and a six-month longitudinal study with 1,000 users. Tiago Ribeiro joins as the first full-time researcher on Zeeguu.
- 2022
- More than 1,000 learners have used Zeeguu, translating over 250,000 words.
- 2021
- The Danish Agency for International Recruitment and Integration funds a University of Copenhagen project that uses Zeeguu to study personalization in Danish courses. Learners pass 200,000 translations.
- 2020
- A high school in Rotterdam adopts Zeeguu. Learners pass 100,000 translations and reach 40,000 distinct words practiced.
- 2019
- A high school in Amsterdam adopts Zeeguu for French.
- 2018
- The University of Groningen Language Centre uses Zeeguu in its Dutch classes.
- 2017
- Gomarus College in the Netherlands starts using Zeeguu in French classes.
- 2016
- Zeeguu goes online for its first beta testers.
People & funding
Zeeguu is led by Mircea Lungu at the IT University of Copenhagen, with Tiago Ribeiro as full-time researcher. Iga Waclawska and Anna Semeriuk joined as researchers on the Villum project. Collaborators include Pernille Hvalsøe (University of Copenhagen) and Nora Hollenstein. Many students have contributed over the years.
Funding: the Villum Foundation (Villum Experiment, 2023) and the Danish Agency for International Recruitment and Integration (2021).
Collaborate
- Teachers and researchers: we're happy to support classroom studies. Get in touch with Mircea.
- Students: see the thesis and project ideas.
- Developers: everything is open source at github.com/zeeguu.
To cite Zeeguu in general, please use As We May Study (CHI 2018):
@inproceedings{lungu2018aswemaystudy,
title = {As We May Study: Towards the Web as a Personalized Language Textbook},
author = {Lungu, Mircea F. and van den Brand, Luc and Chirtoaca, Dan and Avagyan, Martin},
booktitle = {Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems},
series = {CHI '18},
year = {2018},
pages = {1--12},
publisher = {ACM},
address = {Montr{\'e}al, QC, Canada},
doi = {10.1145/3173574.3173912}
}