CER
The most commonly used accuracy metric in OCR and speech recognition, computed by dividing the total number of substituted, deleted, and inserted erroneous characters between the model output and the reference answer by the total number of reference characters. A lower value indicates more accurate transcription.
Core Facts
Timeline (last 90 days)
OCR commonly uses character error rate (CER) and word error rate (WER) as evaluation metrics; document AI focuses more on field-level accuracy and the TEDS score for table reconstruction.
Character Error Rate (CER) is defined as the edit distance (insertions + deletions + substitutions) divided by the total number of characters in the reference text; Word Error Rate (WER) is analogous but operates on words.
CER is more commonly used for languages without spaces such as Chinese, while WER is better suited to English.
The main scoring metric is the paired Nastaliq CER minus Naskh CER, where a positive value means the same source string produced more character errors in Nastaliq.
CER = (S + D + I) / N, where S is substitutions, D is deletions, I is insertions, and N is the total number of characters in the reference answer.
Traditional document parsing evaluations commonly use CER (Character Error Rate) or BLEU scores, metrics that are statistically reasonable but difficult to map to business-relevant questions.
Academia typically measures OCR quality using Character Error Rate (CER) and Word Error Rate (WER), and these metrics have already approached human-level performance on standard printed-text datasets.
OCR accuracy is typically measured using two metrics: Character Error Rate (CER) and Word Error Rate (WER)
Character Error Rate (CER) uses the same calculation logic as WER but operates at the character level, making it more suitable for languages such as Chinese and Japanese that lack clear word boundaries.
Commonly used evaluation metrics in the OCR field include character-level accuracy, word-level accuracy, and Character Error Rate (CER) based on edit distance.
All Facts (10)
OCR accuracy is typically measured using two metrics: Character Error Rate (CER) and Word Error Rate (WER)
75%UnverifiedOCR commonly uses character error rate (CER) and word error rate (WER) as evaluation metrics; document AI focuses more on field-level accuracy and the TEDS score for table reconstruction.
50%UnverifiedCharacter Error Rate (CER) is defined as the edit distance (insertions + deletions + substitutions) divided by the total number of characters in the reference text; Word Error Rate (WER) is analogous but operates on words.
50%UnverifiedCER is more commonly used for languages without spaces such as Chinese, while WER is better suited to English.
50%UnverifiedThe main scoring metric is the paired Nastaliq CER minus Naskh CER, where a positive value means the same source string produced more character errors in Nastaliq.
50%UnverifiedCER = (S + D + I) / N, where S is substitutions, D is deletions, I is insertions, and N is the total number of characters in the reference answer.
50%UnverifiedTraditional document parsing evaluations commonly use CER (Character Error Rate) or BLEU scores, metrics that are statistically reasonable but difficult to map to business-relevant questions.
50%UnverifiedAcademia typically measures OCR quality using Character Error Rate (CER) and Word Error Rate (WER), and these metrics have already approached human-level performance on standard printed-text datasets.
50%UnverifiedCharacter Error Rate (CER) uses the same calculation logic as WER but operates at the character level, making it more suitable for languages such as Chinese and Japanese that lack clear word boundaries.
50%UnverifiedCommonly used evaluation metrics in the OCR field include character-level accuracy, word-level accuracy, and Character Error Rate (CER) based on edit distance.
50%