Discourse Topics and Metaphors

Using metaphor-annotated material that is sufficiently representative of the topical composition of a similar-length document in a large background corpus, we show that words expressing a discourse-wide topic of discussion are less likely to be metaphorical than other words in a document. Our results suggest that to harvest metaphors more effectively, one is advised to consider words that do not represent a discourse topic. Traditionally, metaphor detectors use the observation that a metaphorically used item creates a local incongruity because there is a violation of a selectional restriction, such as providing a non-vehicle object to the verb derail in Protesters derailed the conference. Current state of art in metaphor detection therefore tends to be “localistic” – the distributional profile of the target word in its immediate grammatical or collocational context in a background corpus or a database like WordNet is used to determine metaphoricity (Mason, 2004; Krishnakumaran and Zhu, 2007; Birke and Sarkar, 2006; Gedigian et al., 2006; Fass, 1991). However, some theories of metaphor postulate certain features of metaphors that connect it to the surrounding text beyond the small grammatical or proximal locality. For example, for Kittay (1987) metaphor is a discourse phenomenon; although the minimal metaphoric unit is a clause, often much larger chunks of text constitute a metaphor. Consider, for example, the TRAIN metaphor in the following excerpt from a Sunday Times article on 20 September 1992: Thatcher warned EC leaders to stop their endless round of summits and take notice of their own people. “There is a fear that the European train will thunder forward, laden with its customary cargo of gravy, towards a destination neither wished for nor understood by electorates. But the train can be stopped,” she said. In the example above, the quotation is not in itself a metaphor, as there is no indication that something other than the actual train is being discussed (and so no local incongruities exist). Only when situated in the context prepared by the first sentence (and indeed the rest of the article), the train imagery becomes a metaphor. According to Kittay, a metaphor occurs when a semantic field is used to discuss a different content domain. The theory therefore predicts that a metaphorically used semantic domain would be offtopic in the given document. Although a single document can have singular, idiosyncratic topics, it is likelier to discuss a mix of topics that are typical of the discourse of which it is part. We therefore derive the following hypothesis: Words in a given document that represent a common topic of discussion in a corpus of relevant documents would be predominantly non-metaphorical. That is, a smaller share of metaphorically used words in a document would fall in such topical words than the share of topical words in the document. We test this hypothesis in the current article.