Rigorously Collecting Commonsense Judgments for Complex Question-Answer Content