A case study of using natural language processing to extract consumer insights from tweets in American cities for public health crises