Towards a Voice-Based Chatbot for Language Learners (ChaLL)
Auch auf Deutsch verfügbar
Content of this page
Profile
- Project leadership
- Luzia Sauer (PHZH), Michael Geiss (PHZH)
- Organisational unit
- F-BW-BDW (Bildung und Digitaler Wandel)
- Project phase
- Completed (1.2.2023 – 31.7.2024)
- Project type
- Research & development project
- Funding
- Innosuisse
- Participating institutions
- Zürcher Hochschule für Angewandte Wissenschaften ZHAW; Universität Zürich
Description
Time to practice speaking English during lessons is often very limited. However, in order to offer pupils the best-possible preparation for an increasingly globalised world – in which communications are largely interactive – conversational speaking should become routine already at the primary school level. The development of a voice-based chatbot has the potential to compensate for the fact that, during the few weekly lessons in the curriculum, many pupils rarely speak in class and approaches to teaching speaking strategies are often poorly structured. With the chatbot, pupils can practise English at school or at home on their own, independently with no time constraints and without fear of being judged, thus enabling them to improve their foreign language skills.
Project outcomes
“Towards a Voice-based Chatbot for Language Learners (ChaLL)” was a collaborative project between ZHAW, PHZH, and UZH. It was funded by Innosuisse and ran for 18 months. The project developed a prototype of an interactive chatbot aimed at providing primary school students with more opportunities to practice spoken English. The project idea emerged from the recognition that speaking practice is often limited in English classes, with few opportunities for students to engage in extended, real-life dialogue with language experts. The prototype enables students to practice English anywhere and anytime with a virtual interlocutor, allowing them to explore topics of their choice and at their own pace. This promotes English skills in a relaxed, step-by-step learning environment.
The chatbot, based on advanced natural language processing technology, is designed to simulate natural dialogue, adapting to the individual language level of each learner. It promotes linguistic development by encouraging students to produce language output, recognizing errors and knowledge gaps, and responding in a way that supports foreign language growth. The system thus mimics the personalized learning support typically provided by a teacher.
To create the chatbot, over a hundred hours of speech data from primary school students were collected, anonymized, and annotated. This data formed the basis of the chatbot, allowing it to retain students’ speech errors to provide targeted feedback. To make learning as varied and realistic as possible, different voices and accents – including children’s voices – were integrated into the system. It also features built-in language support, including formulaic chunks and a dictionary.
The chatbot currently offers two prototype communication modes: Chit Chat, where students can speak freely about topics of their choice to build fluency, and Challenges, which focus on specific grammatical structures and vocabulary within a communicative context. The Challenges align with task-based learning principles and the Swiss curriculum, Lehrplan 21, facilitating classroom integration and targeting specific learning objectives.
Initial school tests yielded positive feedback and provided valuable data for further development. Analysis of user experience offered insights that will help in tailoring the chatbot to the needs of young learners.
In summary, the ChaLL project and its voice-based chatbot lay the groundwork for making English learning at the primary level more interactive, motivating, independent, and authentic. While the prototype is ready for use, further optimization is planned to enhance user-friendliness and refine specific learning activities. The ultimate aim is to integrate authentic and sustainable language practice into everyday school life, preparing students for a globally connected world.
Conference Papers
Luzia Sauer & Margarida Pereira: “Developing ChaLL: Towards a voice-based chatbot for language learners”. Oral presentation at Evidence-based foreign language learning – Multilingualism in education. Ascona, 7–10 June 2023.
Manuela Hürlimann, Luzia Sauer, Nicola Good, Gerold Schneider & Mark Cieliebak: “The ChaLL Project: Developing a Voice Bot for Practising a Foreign Language”. Poster Presentation at 8th Swiss Text Analytics Conference. Neuchâtel, 12–14 June 2023.
Luzia Sauer & Manuela Hürlimann: “Developing ChaLL: A digital speaking companion for language learners”. Oral presentation at Language Learning and Teaching in Digital Transformation. Lucerne, 1–2 September 2023.
Manuela Hürlimann, Luzia Sauer, Gerold Schneider, Johannes Graën, Jean-Philippe Goldman, Janick Michot, Katsiaryna Mlynchyk, Ahmet Yavuz Uluslu, Irina-Camelia Stroescu, Jan Deriu, Michael Geiss, and Mark Cieliebak: “ChaLL – A Chatbot for Language Learners”. Poster presented at the SwissText Analytics Conference. Chur, 10–11 June 2024.
Pius von Däniken, Janick Michot, Manuela Hürlimann, Jan Deriu, Luzia Sauer, Katsiaryna Mlynchyk, Mark Cieliebak: Error-preserving Automatic Speech Recognition of Young English Learners’ Language. Poster presented at The 62nd Annual Meeting of the Association for Computational Linguistics in Thailand, 11–16 August 2024.
Luzia Sauer, Manuela Hürlimann, Katsiaryna Mlynchyk, Gerold Schneider, Mark Cieliebak, Michael Geiss, Jean-Philippe Goldmann: “Boosting English Speaking Skills with a Chatbot for Children”. 2nd International Education Conference. Berlin, 9–11 August 2024.
Gerold Schneider, Johannes Graën, Manuela Hürlimann, Luzia Sauer, Janick Michot, Mark Cieliebak, Matthias Bachann: “A Voice-Based Chatbot for Language Learners”. ICAME 2024 Conference, Vigo, 22 June 2024.
Publications
Michot, Janick, Manuela Hürlimann, Jan Deriu, Luzia Sauer, Katsiaryna Mlynchyk, and Mark Cieliebak (2024) Error-preserving Automatic Speech Recognition of Young English Learners’ Language.