Omani-Chatbot / documentation /wer_report.md
Russell Jeffrey
Files
9532ac2

A newer version of the Streamlit SDK is available: 1.52.1

Upgrade

Word Error Rate (WER) Report


This document presents the Word Error Rate (WER) evaluation results for the speech-to-text pipeline. We carried out five separate WER tests. The evaluation yielded the following results:

  • WER for audio_sample_1.wav vs transcript1.txt: 0.0940
  • WER for audio_sample_2.wav vs transcript2.txt: 0.1129
  • WER for audio_sample_3.wav vs transcript3.txt: 0.0958
  • WER for audio_sample_4.wav vs transcript4.txt: 0.0599
  • WER for audio_sample_5.wav vs transcript5.txt: 0.1841