External Captioning Quality Audit - Ai-Media

Quarter ended - Year	Number of programs sampled	NER score Average
Jun-20	13	99.5
Mar-20	13	99.6
Dec-19	13	99.5
Sep-19	13	99.3
Jun-19	13	99.4
Mar-19	13	99.4
Dec-18	13	99.5
Sep-18	13	99.5
Jun-18	13	99.5
Mar-18	13	99.3
Dec-17	13	99.5
Sep-17	13	99.5
Jun-17	13	99.4
Mar-17	13	99.6
Dec-16	13	99.6
Sep-16	13	99.4
Jun-16	13	99.5
Mar-16	13	99.6
Dec-15	13	99.6
Sep-15	13	99.3
Jun-15	13	99.4
Mar-15	13	99.3
Dec-14	13	99.3
Sep-14	13	99.2
Jun-14	13	99.5
Mar-14	13	99.1

*The NER model is a method of calculating the accuracy (quality) of live captions. NER stands for; Number, Edition error, and Recognition error.

A NER value of 100 indicates that the content was captioned entirely correctly.

The NER score can be calculated using the following formula:

N (Number) = Total number of words in the live captions.

E (Edition error) = Edition error

R (Recognition error) = Recognition error