We can not solidly rely on our AI component only. It is good to have a feedback mechanism to evaluate the result of our AI mechanism after a specific time-frame. When we get fresh news to investigate and find the answer to a fundamental question for that news, normally we need to wait for a couple of days (or week's maybe) for the truth to be revealed.
Then we can compare our AI results with the truth and find out how accurate our application is. After collecting the feedback, later we can decide what to do with it. We can tweak our current business log to achieve better results, for example, or we can add new pieces of logic and new mechanisms to catch all the missing parts and improve the quality of our application.
So the wire-frame for the Accuracy Manager is very simple: it looks for the same news for the given keywords after a specific time frame. If the new version of that news exists it notifies us and fetches a fresh copy of it. Then we can compare...