This is a Pyannote database configuration for Spjallrómur. It includes a train, development and test split of the data as well as a single conversation gold standard annotation.
ATT! This is still a work in progress.
Download Spjallrómur from Clarin and unzip in this location.
You can use the following curl command:
curl --remote-name-all https://repository.clarin.is/repository/xmlui/bitstream/handle/20.500.12537/187{/spjallromur_README.txt,/spjallromur_full_conversations.zip}
Load Spjallrómur with pyannote:
from pyannote.database import registry, FileFinder
registry.load_database("data/Spjallromur-pyannote/database.yml")
dataset = registry.get_protocol(
"Spjallromur.SpeakerDiarization.words_and_vocal_sounds",
preprocessors={"audio": FileFinder()}
)