Word Error Rate
A standardized evaluation metric for measuring the accuracy of speech recognition and text transcription systems. It is calculated as the sum of substitutions, insertions, and deletions divided by the total number of words in the reference transcript; lower values indicate higher transcription quality.
Core Facts
Timeline (last 90 days)
WER is suitable as a baseline metric for regression testing but cannot serve as acceptance criteria on its own.
Word error rate (WER) is defined as (substitutions + deletions + insertions) / number of reference words, with lower values indicating closer match to the reference text.
Public word error rate (WER) figures come from benchmark datasets and cannot represent accuracy under a user's own accent, microphone, or industry terminology.
Word error rate (WER) is the core metric for measuring speech transcription accuracy, calculated as the sum of substitutions, deletions and insertions divided by the total number of words in the reference text.
For scenarios such as meeting minutes, a WER exceeding 15% may severely affect readability.
WER is typically measured on standard test sets with clear pronunciation and quiet backgrounds, and real-world performance in actual meeting environments is often worse than official figures.
Word Error Rate (WER) is calculated by dividing the sum of substitution, deletion, and insertion errors between the recognition output and the reference text by the total number of words in the reference text.
Earnings22 is a benchmark dataset specifically designed for speech recognition evaluation in the financial domain, consisting of real earnings call recordings from listed companies, containing a large number of industry terms, numerical expressions, and non-native accents, with word error rates generally higher than in general-domain scenarios.
In scenarios where regularization is needed, mini-KFAC EWC can achieve lower word error rates and a more stable optimization process compared to standard diagonal EWC.
Models fine-tuned on MSA corpora experience a sharp spike in word error rate when encountering real spoken language.
2 more timeline events
All Facts (12)
Word Error Rate (WER) is calculated by dividing the sum of substitution, deletion, and insertion errors between the recognition output and the reference text by the total number of words in the reference text.
80%UnverifiedWER is suitable as a baseline metric for regression testing but cannot serve as acceptance criteria on its own.
50%UnverifiedWord error rate (WER) is defined as (substitutions + deletions + insertions) / number of reference words, with lower values indicating closer match to the reference text.
50%UnverifiedPublic word error rate (WER) figures come from benchmark datasets and cannot represent accuracy under a user's own accent, microphone, or industry terminology.
50%UnverifiedWord error rate (WER) is the core metric for measuring speech transcription accuracy, calculated as the sum of substitutions, deletions and insertions divided by the total number of words in the reference text.
50%UnverifiedFor scenarios such as meeting minutes, a WER exceeding 15% may severely affect readability.
50%UnverifiedWER is typically measured on standard test sets with clear pronunciation and quiet backgrounds, and real-world performance in actual meeting environments is often worse than official figures.
50%UnverifiedEarnings22 is a benchmark dataset specifically designed for speech recognition evaluation in the financial domain, consisting of real earnings call recordings from listed companies, containing a large number of industry terms, numerical expressions, and non-native accents, with word error rates generally higher than in general-domain scenarios.
50%UnverifiedIn scenarios where regularization is needed, mini-KFAC EWC can achieve lower word error rates and a more stable optimization process compared to standard diagonal EWC.
50%UnverifiedModels fine-tuned on MSA corpora experience a sharp spike in word error rate when encountering real spoken language.
50%UnverifiedWord error rate measures the proportion of edit distance between a model's output and the ground-truth transcription, and is the most core accuracy metric.
50%Unverified对于追求智能而非逐字照搬的转写系统，可能需要引入语义相似度（如 BERTScore）、可读性评分、信息完整度等更综合的评估维度
50%