AUDIODIALOGUE
Natural English Dialogue Sets
Natural multi-speaker conversation with the turn structure, timing and speaker context needed for ASR, diarisation and conversational systems.
View datasetAUDIOACOUSTIC SCENES
Acoustic Environment Sets
Ambient recordings are much more useful when the conditions that produced them remain visible. Acoustic Environment Sets organise audio around the scene, device and environmental context that matter to the modelling task.
Where individual sound events are important, they can also be represented on the timeline so scene-level context and event-level information coexist in the same collection. This allows a model to work with both the broader acoustic environment and the specific sounds occurring within it.
Technical structure
- Ambient audio linked to acoustic-scene or environment metadata
- Recording device and condition information where relevant
- Speech-presence or other scene attributes where useful
- Timed sound-event annotations when event localisation is part of the task
Typical applications
Acoustic scene classification, sound-event detection, environmental audio modelling and speech-system robustness.
Delivery note
Scene taxonomy, recording conditions and event granularity are defined around the target acoustic problem.
Delivery
The exact package depends on the collection scope. Where relevant, delivery can include task-specific records, manifests, provenance fields, stable identifiers, SHA-256 hashes, MLCommons Croissant 1.0 metadata and loading instructions.