British Sign Language Glossaries of Curriculum Terms

BSL AI Glossary - Accuracy definition

Translation: Accuracy in AI means how often an AI system gets the right answer or makes the correct decision. Imagine you ask AI to caption a video. If it gets 9 out of 10 words right - that's pretty accurate. If it gets 5 out of 10 right - that's not accurate enough to be useful. Accuracy is basically AI's score for getting things right.

Why AI isn't always accurate, it depends on training data. AI learns from examples. If it was only trained on certain types of data - like one accent, or one language - it will struggle with anything different. It only knows what it has been taught. Auto-captions on YouTube or Teams were mostly trained on clear, standard speech from hearing people. So they often struggle with strong accents, background noise or unfamiliar names and scientific words. That's an accuracy problem - and it directly affects the captioning.

Another example, if you are using an AI map and it says your journey will take 31 minutes, but it actually takes 50 minutes - that is low accuracy. But if you use a different AI map system, and it says 31 minutes and you arrive in exactly 31 minutes - that is high accuracy. Why accuracy is important, in everyday apps, low accuracy is annoying. But in serious situations - like AI used in healthcare - low accuracy can be really harmful.

AI accuracy means how often it gets things right. But AI is not perfect - and it depends on the training data.