🎙 Multilingual AI Scribe Accuracy Simulator

The Visit Starts in the Patient's Language

Placeholder: ambient scribe listens in from the very first spoken word.

• 9: Languages modeled (placeholder short caption)
• 3: Resource tiers (placeholder short caption)
• Ambient audio: Input mode (placeholder short caption)
• 2: Downstream steps (placeholder short caption)

Placeholder section heading text

Placeholder body text, short and generic for now.

Speech-to-Text Accuracy Tracks Training Data Volume

Placeholder: word-level accuracy differs sharply by language resource level.

• ~97%: High-resource accuracy (placeholder short caption)
• ~88%: Medium-resource accuracy (placeholder short caption)
• ~76%: Low-resource accuracy (placeholder short caption)
• up to 18pt: Accent penalty (placeholder short caption)

Placeholder section heading text

Placeholder body text, short and generic for now.

Clinical Vocabulary Capture Lags Behind Plain Speech

Placeholder: medical terms are harder to capture than everyday words.

• ~95%: High-resource term acc. (placeholder short caption)
• ~82%: Medium-resource term acc. (placeholder short caption)
• ~65%: Low-resource term acc. (placeholder short caption)
• widens downstream: Gap vs. transcription (placeholder short caption)

Placeholder section heading text

Placeholder body text, short and generic for now.

Side-by-Side Accuracy Across Resource Tiers

Placeholder: the gap becomes visible once tiers sit side by side.

• ~96%: Best tier combined acc. (placeholder short caption)
• ~70%: Worst tier combined acc. (placeholder short caption)
• ~25pt: Tier spread (placeholder short caption)
• 9: Languages compared (placeholder short caption)

Placeholder section heading text

Placeholder body text, short and generic for now.

Accuracy Disparities Raise Real Equity Concerns

Placeholder: lower accuracy means more physician review is needed.

• ≥90% combined: Standard review threshold (placeholder short caption)
• <90% combined: Enhanced review threshold (placeholder short caption)
• Lower-resource speakers: Populations affected (placeholder short caption)
• Targeted physician review: Mitigation (placeholder short caption)

Placeholder section heading text

Placeholder body text, short and generic for now.

Placeholder highlight: equity requires monitoring accuracy per language.