Hum Rahi

User Story 4 (BE/AI): Voice Note Transcription & NLP Analysis¶

Story: As a Backend System, I want to transcribe voice notes and analyze the text, so that contextual details are extracted from audio reports.

• In-Scope:
  • Speech-to-Text (STT) conversion.
  • Keyword extraction (e.g., "weapon," "bleeding," "trapped").
  • Updating the report record with the transcript.
• Out-of-Scope: Language translation (focus on English/Local primary).
• Acceptance Criteria (AC):
  1. Audio notes must be transcribed with >80% word accuracy.
  2. Transcript must be searchable in the database.
  3. Detected keywords must influence the Severity Level (US-020).
• Validation Rules:
  • Audio length: 2s to 30s.
• Error Messages:
  • "Audio processing failed: Excess background noise."
• DoD: STT latency <3 seconds for 30s audio, transcription storage verified.