CAIV Methodology and Process
CAIV studies are based on a modified WER (Word Error Rate) methodology which provides an objective percentage of word accuracy based on total word errors and omissions, taking context, formatting, and punctuation out of the accuracy calculation. This makes for a more empirical data-driven methodology compared to the other methods of speech-to-text analysis. To eliminate the subjectivity of punctuation and speaker identification inherent in AI platforms and live human captioning, CAIV can provide a separate score for speaker identification and/or punctuation accuracy, viewable alongside our WER based percentage score.
Contact Us
