WER is a Statistical Metric
Input
- Required Inputs:
actual_output: The Generated Transcripts given by the modelexpected_output: Reference transcript text
Output
Result: WER percentage (0–100, lower is better)Reasoning: Brief breakdown of errors (S, D, I) and overall accuracy
Interpretation
- 0%: Perfect transcription with no errors
- Higher percentages: More transcription errors, lower accuracy