Using sem-grams to study metaphors in travel journalism

The current paper introduces the corpus-driven technique of the “sem-gram”, on analogy with the n-gram and the PoS-gram. As the latter, it is essentially a concatenation of tags, but while the PoS-gram is a string of Part-of-Speech tags, the sem-gram is formed of a blend of PoS and semantic tags, the latter being used when the token in question is PoS-tagged as being a lexical verb, an adjective or a common noun. The usefulness of the sem-gram technique will be illustrated by way of the analysis of a 4M token corpus of travel journalism from the online version of the British daily 'The Guardian'. The procedure allowed the automatic identification of metaphorical expressions from the source domain of the human body.