Dictate report content using speech-to-text after a session
Peer mentors, particularly those supporting people with visual impairments or those who are older and less comfortable with touch keyboards, need the ability to dictate report content rather than type it. The speech-to-text overlay is activated per field and transcribes speech in real time into the report form. Recording only happens in the reporting phase — never during the actual session — to protect the privacy of the contact. Partial transcripts are shown in a preview field so the peer mentor can review and correct before accepting.
User Story
Audience Summaries
Speech-to-text report dictation directly addresses accessibility barriers for peer mentors supporting users with visual impairments or low digital literacy — a critical segment for partner organisations Blindeforbundet and HLF. By enabling natural voice input during the post-session reporting phase, the system removes the single largest friction point preventing timely, detailed report submission. This translates to measurably higher report completion rates and richer data quality for Bufdir compliance reporting, reducing underreporting — identified as the top pain point across all partner organisations. The feature also signals commitment to inclusive design, strengthening credibility with disability-focused funders and expanding the platform's suitability for a broader range of peer mentor profiles.
Retention of peer mentors who would otherwise disengage due to reporting burden is a direct cost-saving outcome.
This story has medium priority but non-trivial delivery complexity due to its dependency on the structured post-session report screen (story-structured-post-session-report-peer-mentor-001) being stable before implementation begins. Cross-functional coordination is required between mobile developers and accessibility/QA specialists to validate real-time transcription performance across supported devices. Acceptance criteria define six distinct test scenarios including microphone unavailability, session-state gating (speech-to-text must not be available during active sessions), and transcript editing — each requiring dedicated test cases. Privacy implications of microphone activation must be reviewed with a legal or compliance stakeholder before release.
Rollout should include a pilot with Blindeforbundet users to validate transcription accuracy in Norwegian. A fallback text input path must remain fully functional throughout, limiting regression risk.
Implementation requires integrating a device-native or cross-platform speech-to-text API (e.g., Web Speech API for web or platform speech recognisers for React Native) with per-field activation logic on the post-session report form. A microphone button component must be added to each text field, triggering an overlay that streams partial transcript previews in real time. State management must track whether the current context is an active session or a completed one — blocking speech-to-text activation until the session is explicitly ended, enforced at the UI layer and validated server-side. The transcript acceptance flow must insert confirmed text into the bound form field without disrupting existing form state.
Error handling is required for microphone permission denial and hardware unavailability, with graceful degradation to standard text input. Edge cases include background noise interference, mid-sentence pauses, and multi-language input. Unit tests should cover state gating logic; integration tests should cover the full transcription-to-field-insert flow.
Acceptance Criteria
- Given a peer mentor is on the post-session report screen, when they tap the microphone button next to a text field, then the speech-to-text overlay activates and begins listening
- Given the speech-to-text overlay is active, when the peer mentor speaks, then a partial transcription preview appears in real time
- Given a transcription is complete, when the peer mentor accepts it, then the transcribed text is inserted into the corresponding report field
- Given the peer mentor is still in an active session context, when they attempt to open a report form, then speech-to-text is not offered until the session is explicitly ended
- Given the device microphone is unavailable, when the peer mentor tries to activate speech-to-text, then a clear error message is shown and the text input remains available
- Given dictated content contains errors, when the peer mentor reviews the transcript preview, then they can edit it before accepting
Business Value
Speech-to-text significantly lowers the barrier to completing reports for peer mentors who support users with disabilities — particularly for Blindeforbundet and HLF. It enables faster, more natural reporting and increases the likelihood of detailed, accurate reports being submitted. This directly supports Bufdir reporting quality and reduces the underreporting problem identified as the top pain point across all organisations.
Components
- Speech-to-Text Field Overlay ui
- Speech-to-Text Adapter infrastructure
- Post-Session Report Screen ui
- Dictation Microphone Button ui
- Recording State Indicator ui
- Transcription Preview Field ui
- Speech Recognition Service service
- Transcription State Manager service
- Dictation Scope Guard service
- Partial Transcription Repository data
- Native Speech API Bridge infrastructure
- Accessibility Live Region Announcer infrastructure