Live captioning accuracy in English‑language newscasts in the USA
After the Federal Communications Commission issued its first closed captioning quality regulations in 2014, the captioning industry in the USA must monitor the closed captions they produce in order to ensure they comply with the rules in place. In the case of live captioning, accuracy becomes a key aspect since, together with other parameters such as speed or latency, it shapes quality. Live captioning accuracy is often measured using the word error rate (WER), an instrument that informs about the words that have been mistakenly inserted, deleted and substituted in a closed caption. While this is convenient because WER is a fully automated metric, it does not consider correct edition or the fact that some captioning errors hamper viewers’ comprehension more than others. In order to account for correct edition and error severity, this article reports on the main findings of a research project aimed at exploring closed captioning accuracy using the NER model. The closed captions accompanying the national newscasts broadcast by four networks in the USA were analyzed for accuracy. The results point at an overall good accuracy, with almost 2/3 of the errors being minor, and slightly over 1/3 being standard or serious.
Fresno, N. Live captioning accuracy in English-language newscasts in the USA. Univ Access Inf Soc (2023). https://doi.org/10.1007/s10209-023-01030-w
Universal Access in the Information Society