Romer (2008) uses a Sinclairian approach to corpus analysis in order to identify positive and negative evaluations in texts such as book reviews: how can we find out quickly and reliably (and preferably automatically, by computer program) whether a reviewer thinks a book is worth bothering with?