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 datasetAUDIOMULTILINGUAL
Cross-Language Dialogue Sets
Multilingual speech becomes difficult to compare when every language arrives with a different structure. Cross-Language Dialogue Sets use a common record model across language groups while leaving room for locale-specific transcript conventions, scripts and linguistic metadata.
That balance matters for teams developing one system across several markets. The underlying relationships between audio, speakers, turns and metadata can stay consistent, while the fields that need to vary by language remain explicit rather than being forced into an unsuitable universal format.
Technical structure
- Shared session and utterance schema across language groups
- Language, locale, dialect or accent metadata where relevant
- Speaker-labelled transcripts with utterance timing
- Optional translation, romanisation or language-specific annotation layers
Typical applications
Multilingual ASR, language identification, voice-agent localisation, speech translation and cross-language model comparison.
Delivery note
Languages and transcript conventions are defined per project so the common structure does not erase meaningful local differences.
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.