Measuring verbal and non-verbal features of L2 learners' spoken interaction: Rethinking automated speaking assessment.

The poster "Measuring verbal and non-verbal features of L2 learners’ spoken interaction: Rethinking automated speaking assessment" was presented at Language Testing Research Colloquium 2024, July 1-5, Innsbruck, Austria.

Abstract 

Automated speaking assessment (Zechner & Evanini 2020) is often limited to individual performance. Moreover, scales used for assessing second language (L2) spoken interaction rarely include non-verbal behaviors such as gaze and gestures or intonational cues (see e.g. Council of Europe 2020), although they are important in non-test conversations.

This poster presents aims and starting points of the Aasis research project (2023–2027), which focuses on verbal and nonverbal features of L2 Finnish learners’ spoken interaction and develop ways to assess these automatically. Aasis builds on a previous project, DigiTala, which developed an ASR-based (automatic speech recognition) online tool for assessing L2 Swedish and Finnish learners’ speech automatically and providing automated feedback to the language learners.

The project aims to expand ASR-based L2 speaking assessment to cover also assessment of interaction skills. In addition to L2 speech, the research interests include non-verbal communication such as body language and interactional phonetics.

The methods include 1) videoing academic L2 Finnish learners’ dialogues, 2) training human raters to assess and transcribers to annotate learners’ performances, 3) analyzing the ratings using Many-facet Rasch measurement and 4) experimenting machine learning methods to predict human ratings. Where possible, the data collected and tools developed are published following the principles of open science.

Automatic assessment of interaction improves authenticity and reliability of spoken L2 assessment by enabling ASR-based dialogue speaking tests and supporting human raters’ and teachers’ work. Moreover, automatic scores produced by the machine could be used for providing automated feedback to the learners.

Acknowledgements

I thank the researchers of the Aasis consortium (Automatic assessment of spoken interaction in a second language, Research Council of Finland 2023–2027) at the University of Helsinki (grant number 355586), Aalto University (grant number 355587) and the University of Jyväskylä (grant number 355588). The current project builds on experiences gained during the consortium’s previous project, DigiTala (Academy of Finland 2019–2023, grant numbers 322619, 322625, 322965). I also express my gratitude for the Helsinki Institute for Social Sciences and Humanities Catalyst Grant 2024, which has made it possible to adopt new methods such as eye-tracking and the computer-joystick method. 

More information

von Zansen, A. (2023). The Aasis research project: automatically assessing spoken interaction in L2 Finnish. Language, Education and Society, 14(7). Available: https://www.kieliverkosto.fi/fi/journals/kieli-koulutus-ja-yhteiskunta-joulukuu-2023/the-aasis-research-project-automatically-assessing-spoken-interaction-in-l2-finnish