Analysis - CER (Character Error Rate) and WER (Word Error Rate)

CER and WER are metrics for evaluating speech recognition and machine translation.

Character Error Rate (CER):

• Measures character-level accuracy.
• Formula: (S + D + I) / N
  • S: Substitutions
  • D: Deletions
  • I: Insertions
  • N: Total characters in the reference

Word Error Rate (WER):

• Measures word-level accuracy.
• Formula: (S + D + I) / N
  • S: Word substitutions
  • D: Word deletions
  • I: Word insertions
  • N: Total words in the reference

Example:

Reference: "the quick brown fox"
Hypothesis: "the quikc brown fox"
WER: 1/4 = 25%
CER: 1/16 = 6.25%

Lower CER and WER values indicate better performance.