NordAlign – Automatic phonetic annotation
of the Nordic languages
LINK | Young, N.J. (2020). NoFA 1.0 – norsk modell for forced alignment, version 1.0. Språkbanken, National Library of Norway. https://www.nb.no/sprakbanken/ressurskatalog/oai-nb-no-sbr-59
Funding body: The Swedish Academy (and self-financed)
NordAlign is the name I have given to language-specific models that I trained on the Montreal Forced Aligner for Danish, Norwegian bokmål, and Swedish, respectively. The software outputs time-stamped phonetic boundaries in Praat with the following inputs: (1) sound files, (2) matching orthographic transcriptions, (3) a language-specific model for the relevant language.
The Swedish model was originally trained on speech material from men, so one ongoing sub-project is collecting speech from women and youth in order to balance the model. Also ongoing is a web app that will allow researchers to access my models in a user-friendly manner. Both of the latter sub-projects are financed by the Swedish Academy. Whereas my Danish and Swedish models are proprietary, the rights to my Norwegian bokmål model are held by the Norwegian National Library and can be accessed here.