Data collection process

How custom datasets are built

From requirement discovery to secure delivery, each collection program is designed around your exact specification.

Requirement discovery

Language, modality, geography, demographic, annotation schema, volume, quality and timeline requirements captured with your team.

Collection design

Collection protocol covering recruitment, recording setup, consent, metadata capture and acceptance criteria.

Recruitment

Participants recruited to match the target demographic, language and speaker profile.

Pilot

Small pilot collection validates quality, consent flow and recording setup before scaling.

Production collection

Full-scale collection under documented guidelines with daily oversight and consent tracking.

Validation

Collected data validated against specification, quality thresholds and metadata completeness.

Metadata

Structured metadata captured per item: language, channel, sample rate, speaker, domain and annotation.

Quality control and delivery

Multi-level QC then secure delivery via SFTP, S3, GCS or controlled access as agreed.

Delivery options

Secure delivery

Request a custom collection

Describe your dataset requirement and we'll propose a program.

Request a collection proposal โ†’