How learning works

Mark extracted events as verified, likely, uncertain or irrelevant on their review page. That feedback adjusts the query usefulness and source trust scores. Productive queries run more often, while exploration prevents the system from becoming trapped in its current assumptions.