The highlighted responses to suggested natural language questions are comically bad. Answers to different questions, complete failure to apply simple contextual

The highlighted responses to suggested natural language questions are comically bad. Answers to different questions, complete failure to apply simple contextualisers like 'today', and treating some random on Quora or some splog as an authoritative source for identifying a fact that has authoritative sources. I know it's difficult to do this stuff without human curation, but it's like a campaign to discourage people from taking NLP seriously.