DigiTala is an interdisciplinary research project that aims to develop a computer-aided oral language test as part of language tests at the end of general upper secondary education. Project partners involve Department of Teacher Education at the University of Helsinki, Phonetics at the Institute of Behavioural Sciences at the University of Helsinki and Department of Signal Processing and Acoustics at Aalto University. We work in co-operation with University of Jyväskylä, Finnish Matriculation Examination Board, The Federation of Foreign Language Teachers in Finland and Finnish National Board of Education.
The project aims at enhancing practicality and reliability of testing speaking in foreign languages by developing, piloting and implementing digital environments and computerized tools for supporting and reducing the workload of human raters. At the same time, the construct of speaking will be scrutinized along with the various quality aspects of a test appropriate for large-scale implementations and high-stakes decision about the candidates’ linguistic skills.
To meet the theoretical and societal goal of the project, the following research questions are addressed:
1. How consistent are the ratings assigned by human raters and the speech recognition tool in regard to fluency, pronunciation, vocabulary range and accuracy of spoken production?
2. How do the ratings compare with other relevant indicators of communicative language ability (e.g. monologue/dialogue tasks, school grades, scores in listening, reading and writing)?
3. What technical and acoustic factors have the greatest impact on the obtained scores (facilities, recording devices, background noise etc.)?
4. What conceptions do students and teachers report regarding computer assisted test tasks and the course of implementation?
The project is financed by the Svenska folkskolans vänner foundation.
DigiTala uses recorded spoken language collected by Society of Swedish Literature in Finland in the project Spara det finlandssvenska talet 2005-2008. The material is available and searchable in the spoken language corpus Talko.