Deterministic First
Real-time language events are delivered predictably and are never blocked by interpretation.
Accessibility Framework
Observable speech, sign, sound, and visual events — deterministic first, enriched later, never overwritten.
Stokoe is a real-time accessibility framework that defines how language and situational information are observed, recorded, and rendered — without collapsing truth into interpretation.
Real-time language events are delivered predictably and are never blocked by interpretation.
Meaning is added in optional layers — never instead of raw events.
Observation, timing, and integrity are preserved at the point of capture.
All events are timestamped, replayable, and auditable.
┌─────────────────────────────────────────────────────────────┐
│ REAL-TIME (Deterministic, Local, Non-blocking) │
├─────────────────────────────────────────────────────────────┤
│ Lane A: Speech-to-Text → Captions (raw transcript) │
│ Lane B: Sound Awareness → Audio events (labeled) │
├─────────────────────────────────────────────────────────────┤
│ ENRICHMENT (Optional, Annotative, Inspectable) │
├─────────────────────────────────────────────────────────────┤
│ Lane C: Speaker Detection → Who said what │
│ Lane D: Semantic Context → Intent, sentiment │
│ Lane E: Knowledge Layer → Linked data, references │
└─────────────────────────────────────────────────────────────┘
↓ Data flows bottom-up (enrichment never blocks A/B)
↓ User controls which lanes are active
↓ Each lane is independently auditable Catalyst is the reference implementation used to observe, test, and validate Stokoe-compliant systems. It demonstrates correct behavior, not optimized presentation.
View Catalyst documentation →