Home

The DANE ASR worker is currently being implemented. For the actual ASR processing Kaldi, or more specifically, Kaldi_NL will be used. The ASR yielded from this software has proven to be of high quality in 2 ASR pipelines that have been implemented priorly. However in their current forms, these pipelines cannot be properly maintained.

Since we now have committed to install & maintain DANE for a variety of content processing use-cases within the x-omgeving, this DANE worker for ASR will follow a proper DevOps approach from the start.

For starters, let's try to make sure Kaldi and Kaldi_NL are easy to install & maintain. With this in mind, the current installation approaches and dependencies for Kaldi are laid out in the following image:

Based on this overview, a good approach for setting up & implementing a new DANE worker for ASR (this repository) should be easier to devise.

TODO