Topic models can produce clusters of words that characterize written documents. But how do we figure out what those clusters mean, exactly?

To capture the degree of information in spoiler review text that reduces plot uncertainty, the authors propose a spoiler intensity metric and measure it using a correlated topic model.