Skip to content

Display Realtime ASR Performance Metrics for Every Recording

Description

The ASR application currently provides realtime transcription output but does not expose processing and quality-related metrics to the user. Add a metrics panel that automatically calculates and displays key ASR performance indicators for each realtime recording.

Metrics to Display

  • Processing Time
  • RTFx (Real-Time Factor)
  • CER (Character Error Rate)
  • WER (Word Error Rate)
  • Detected Language(s)
  • Number of Speakers Detected

RTFx should be calculated as:

RTFx = \frac{\text{Processing Time}}{\text{Audio Duration}}

Expected Behavior

  • Metrics should update automatically after or during realtime transcription.
  • Processing time should reflect total ASR inference duration.
  • RTFx should accurately represent realtime processing efficiency.
  • CER and WER should be displayed when reference transcripts are available.
  • Detected languages should be listed clearly.
  • Speaker count should be obtained from diarization results.
  • Metrics should be displayed in a dedicated UI section/card for each recording session.
  • UI updates should occur dynamically without requiring a page refresh.

Acceptance Criteria

  • All required metrics are visible for every realtime recording.
  • RTFx values are calculated correctly.
  • Language detection and speaker count display accurate results.
  • CER/WER handling works correctly when ground-truth transcripts exist.
  • Metrics UI is responsive and integrated with the existing ASR interface.