Time-Aligned Training Data for Speaker Diarization

It is the labeling of audio intervals to indicate which speaker spoke when. The speaker labels normally identify distinct voices within the recording, such as Speaker A and Speaker B, rather than a person's real-world identity.

Yes. Define the speaker classes or properties and reuse them across the relevant intervals. Reviewers can inspect turn consistency throughout the file.

Create the relevant time intervals for the speakers heard during the overlap, following the dataset policy. Review exact boundaries and short interruptions against the recording.

Yes. Use role-based labels or properties where the recording provides that information. Keep role attribution and speaker identity rules explicit in the ontology.

Yes. Supported visible timed speaker annotations can be exported in RTTM with their recording identity, start time, duration, channel context, and speaker label.

Define shared classes, structured properties, and clear labeling instructions before work starts. Use representative examples and contextual review to resolve disagreements in the speaker diarization dataset.

Yes. Route audio annotation through Review and Rework stages. Reviewers can inspect the source data, correct labels, and send an item back when more work is needed.

Yes. Use dataset search, metadata, tags, and available filters to select relevant audio assets. Keep representative conditions and difficult examples visible in the preparation workflow.

Export reviewed audio labels in supported Audio JSON, RTTM, or UUEF formats according to the annotation task. Dataset versions help teams identify which prepared examples belong to a release.