Information for research participants

The Academy of Finland is funding the DigiTala project 2019-2023, which is conducted in collaboration between the University of Helsinki, the University of Jyväskylä and Aalto University (helsinki.fi/en/digitala). The leaders responsible for the research project are Assistant Professor Raili Hildén from the University of Helsinki (collaboration with schools and collaborators), University Lecturer Mikko Kuronen from the University of Jyväskylä (speech analysis and training for human raters) and Professor Mikko Kurimo from Aalto University (automatic speech recognition and automatic assessment). 

The project develops applications that use automatic speech recognition (ASR) in order to evaluate language skills

The purpose is to investigate how oral language skills in foreign languages ​​can be assessed in large-scale and high-stakes language tests such as the Finnish Matriculation Examination. The applications also allow students to practise pronunciation and speaking on their own. The study focuses in particular on the assessment of pronunciation, vocabulary, grammar and fluency in second and foreign languages. 

The first versions for Swedish or Finnish as a second language

The first versions of the applications are intended for Finnish upper secondary students who study Swedish or Finnish as a second national language. Later, applications may be developed to evaluate other foreign languages. It is important that as many students as possible participate in the research, as the machine itself cannot evaluate students speech, but must be taught to do so. 

The machine needs many speech samples from different students in order to learn

 Speech samples are not sufficient, but comments from teachers and human raters are also required. Using the students speech samples and human ratings, the machine learns to better recognise what the students are saying and how people rate different speech samples and speech samples at different levels.

Research on user experience and fairness of automatic scoring

We also test the automated feedback provided by the machine. It is important for us to know whether the assessments given by the machine are fair to teachers and students. We also study, for example, whether the feedback received from the machine is perceived as useful and understandable. Towards the end of the project, we will also study the usability of the tool.

The students names, dates of birth and schools are removed from the research material before the analysis phase. The privacy notice is available on DigiTala’s website. 

For more information on the research, please contact: Raili Hildén, +358504482514, raili.hilden@helsinki.fi  

Information for upper secondary school students

We can request permission to use the material for the following purposes; you can decide which of these purposes you allow: 

  • We can request permission for a follow-up study where we later compare the students’ grades in the ability to speak with their grades in other language subskills (such as writing, listening and reading) in the Matriculation Examination. This comparison gives us valuable feedback on how well the application we develop works and on the connection between the subskills. 
  • We can request permission to contact you later on during the research when we collect further feedback and development suggestions from students, teachers and raters. We also interview some users. 
  • We can request permission to use the collected speech samples to train language teachers and raters. This requires that some recordings are published as examples. However, we will not publish names, dates of birth or schools with the recordings. 
  • We can request permission to store the material in the Language Bank of Finland. The material collected in this research is also useful in other research related to language learning. Identifiers such as names, dates of birth and schools will be removed from the material stored in the Language Bank. The material in the Language Bank will only be available for legitimate research purposes.