When conducting research, interview data (audio) is collected using the MyJYU AI Transcription mobile application, developed by the University of Jyväskylä for Apple and Android operating systems and based on open-source technology. The application incorporates advanced AI-based methods for automatically transcribing interviews and translating the data into English when needed. This combination of front- and backend technologies expand data collection (number of interviews), reliable collection and storage, and automatic AI-based preliminary transcription, supporting the scientific quality of the research study. The mobile application's usability and integration with the university's data center backend systems ensure that data collection is easy for researchers and secure for participants.
The AI-based transcription of the collected data is performed locally in a research-dedicated server cluster (rc.jyu.fi) located at the University of Jyväskylä’s data center, which is optimized for the AI-based GPU processing and analysis of large data sets. This infrastructure ensures a high level of performance, data security, data integrity, and reliable lifecycle management, making it unique whole solution and particularly suitable for handling sensitive research data. All collected interview data, including original audio recordings, transcripts, and translations, are securely encrypted using modern encryption methods, which protect the data throughout the research process and ensure participant privacy. Authentication into the backend service (rc.jyu.fi) requires strong identity verification and two-factor authentication.
The server cluster (rc.jyu.fi) leverages modern hyperautomation, which manages the metadata and lifecycle of all collected research data, contributing to compliance with the research data management policy. The transcription process utilizes the OpenAI WhisperX Large model, based on advanced machine learning algorithms. This model can recognize and transcribe various speech styles and accents and automatically distinguish between speakers within the interview in up to 90 different languages.
The transcription results are reviewed and, if necessary, manually corrected by the research team outlined in this research plan to ensure that the data is scientifically reliable and of high quality. The server environment includes tools for reviewing and correcting the interview transcriptions. This process is essential for ensuring the quality of data analysis and conclusions. AI-based transcription accelerates the early phases of the research and frees up resources for more in-depth analysis and interpretation of results, making the research process more efficient.
The entire solution is designed specifically for the collection and handling of sensitive research data and meets strict data security and privacy requirements. The mobile application and backend service have undergone a security audit. Reliable and modern research process that produces high-quality results and enhances the impact of the research both nationally and internationally.
MyJYU AI Transcription and background service do not incur additional costs for the researcher or research project. Solution is included as part of the basic digital services provided for researchers at the University of Jyväskylä.