Your work will be part of the OH-SMArt project, a long term initiative to significantly improve the digital research chain around using Oral History and spoken narratives. Automatic Speech Recognition (ASR) is part of this chain.
We are seeking a junior researcher for a full-time 17 months-long position who will be responsible for the implementation and evaluation of an open-source Dutch ASR end-to-end system such as Wav2Vec or Whisper in the CLARIAH research infrastructure. Speech recognition services in the infrastructure are docker based and should be able to run on computer clusters at member organizations (e.g., Netherlands Institute for Sound and Vision, DANS, SURF) to enable processing of data sets on a large scale. Working together with teams at these organizations will be part of the work.
Your tasks will involve:
- Implementation of open-source end-to-end speech technology for Oral History in an existing infrastructure
- Evaluation of the implemented ASR system and comparison with current systems (Kaldi based).
- Optionally, fine-tuning of system performance in collaboration with currently running projects on Dutch ASR in the Netherlands.
For the implementation of the ASR, you will work together with speech technology researchers from Radboud University (Nijmegen) and CLARIAH development teams at Netherlands Institute for Sound and Vision.