Guide
Audio Collection SOP — recruitment to validation
The standard operating procedure for Cognegica audio collection: speaker recruitment diversity, recording-environment standards, device diversity, speech content types, metadata and multi-stage validation.
This SOP documents how Cognegica runs audio collection end to end, from recruitment to validation, so deliveries are balanced, documented and auditable.
1. Speaker recruitment diversity
Sampled across age, gender, regional accents, socio-linguistic backgrounds and recording environments to a defined matrix.
2. Recording-environment standards
Quiet indoor capture, minimal background noise, clear microphone placement and stable connectivity, with controlled noise variations only where the use case requires them.
3. Equipment and device diversity
Android and iOS smartphones, laptop and desktop microphones, headset mics and field recording devices.
4. Speech content types
Scripted prompts, spontaneous speech, scenario-based dialogues and prompt-based responses, each tagged.
5. Metadata collection
Language and dialect, speaker demographics, device type, environment category and location, recorded per record.
6. Multi-stage validation
Automated quality checks, manual review, script-adherence and noise/clarity assessment, and metadata verification; failed recordings are flagged for correction or replacement.
Using this SOP
Questions about audio collection
- How is the corpus kept balanced?
Recruitment is sampled to a defined demographic and environment matrix, and metadata is captured per record so the dataset can be balanced and audited.
- Do you collect spontaneous speech?
Yes — scripted prompts, spontaneous speech, scenario-based dialogues and prompt-based responses are all in scope, tagged by content type.
- What happens to failed recordings?
Multi-stage validation flags them for correction or replacement before delivery.