Use of ELFA corpus for research – Privacy Notice

About this notice

This privacy notice describes how the Aalto ASR research group (Aalto University) uses the ELFA corpus in research on second language (L2) speech and speech technology. This notice does not replace or modify the official terms, licences or privacy information provided by the ELFA corpus and Kielipankki.

Purpose of processing

The ELFA corpus is used as a research dataset to support our work on developing and evaluating methods for analysing and modelling L2 speech and spoken interaction (for example, automatic speech recognition and related speech-technology models).

We do not use the ELFA corpus to assess, evaluate, or profile the individual speakers in the corpus, and we do not use it for any commercial purposes. The corpus is used solely for scientific research.

What data is processed

Depending on the specific experiment, we may process:

• Audio recordings from the ELFA corpus.
• Associated transcripts and annotations provided as part of the corpus.
• Derived features (e.g. acoustic features, model outputs, scores) computed from the audio and transcripts.

We do not add any new personal identifiers to the ELFA data, and we respect the anonymisation and usage conditions defined in the corpus documentation.

Legal basis and data sources

The data is obtained from the ELFA corpus distributed via Kielipankki under its applicable licence and terms of use. The legal basis for processing is the performance of a task carried out in the public interest (scientific research conducted by Aalto University).

Who has access to the data

Access to the ELFA data and derived research data is restricted to authorised members of the research team who have agreed to follow the ELFA and Kielipankki terms of use and Aalto University’s data protection and information security guidelines.

Data sharing and transfers

We do not share the original ELFA data outside the terms defined by Kielipankki. Research results may be published in aggregated or anonymised form (e.g., model performance metrics, model weights), in such a way that individual speakers are not identifiable.

Retention

The ELFA data is stored and retained in accordance with the ELFA and Kielipankki licence terms and Aalto University’s research data management practices. Derived research data (e.g. model outputs) may be stored for as long as necessary to document and reproduce the research.

Your rights

For information about rights related to the ELFA corpus and Kielipankki services, please refer to the official information provided by Kielipankki and the corpus maintainers. If you have questions about how the Aalto ASR research group uses the ELFA corpus in this specific research project, you can contact us using the details below.

Contact

For questions about this privacy notice or the research project, please contact:
Mikko Kurimo, Full Professor, Aalto University
Email: mikko.kurimo@aalto.fi