The 2024
State of ASR
This report will:
- Explore nuances of error rate measurement
- Compare accuracy across industry use cases and transcription styles
- Enable more informed decision-making when selecting/using ASR engines
700 video files
158 hours of content
1.3 million words
Download the report:
3Play's post-processing was shown to
relatively improve error rates by as much as
10%
Word Error Rate
Common metric to measure how much content an ASR engine recognizes, without the inclusion of formatting considerations.
Formatted Error Rate
Evaluates overall experience, readability, and accuracy including elements like punctuation, capitalization, grammar, and other notations.
Other error types
Substitution, deletion, and inclusion errors were also measured in combination with FER and WER to make up the total error rate.
"It has become clear that not all errors are equal, challenging the standalone metric of accuracy rate."
Josh Miller, co-CEO and co-founder
3Play Media
THE LOWER THE BETTER
Word Error Rate (WER) provides a useful benchmark for understanding accuracy rates, although it should ideally be evaluated alongside other important accuracy indicators.
This year, we saw a range in WER performance across engines.
"The ASR market continues to evolve and is fiercely competitive. It's clearly reaching a maturation stage."
Josh Miller, co-CEO and co-founder
3Play Media
A Full Service Media Accessibility Solution
3Play Media offers an integrated platform with patented solutions for closed captioning, transcription, live captioning, audio description, and localization. As a thought leader in video accessibility, we are committed to providing free, educational resources like this one.