Assessing the Potential of Metaphoricity of verbs using corpus data

Assessing the Potential of Metaphoricity of Verbs Using Corpus Data Marco Del Tredici, N´uria Bel Rafecas Universitat Pompeu Fabra Roc Boronat, 138, Barcelona, Spain marco.deltredici@upf.edu, nuria.bel@upf.edu Abstract The paper investigates the relation between metaphoricity and distributional characteristics of verbs, introducing POM, a corpus-derived index that can be used to define the upper bound of metaphoricity of any expression in which a given verb occurs. The work moves from the observation that while some verbs can be used to create highly metaphoric expressions, others can not. We conjecture that this fact is related to the number of contexts in which a verb occurs and to the frequency of each context. This intuition is modelled by introducing a method in which each context of a verb in a corpus is assigned a vector representation, and a clustering algorithm is employed to identify similar contexts. Eventually, the Standard Deviation of the relative frequency values of the clusters is computed and taken as the POM of the target verb. We tested POM in two experimental settings obtaining values of accuracy of 84% and 92%. Since we are convinced, along with (Shutova, 2015), that metaphor detection systems should be concerned only with the identification of highly metaphoric expressions, we believe that POM could be profitably employed by these systems to a priori exclude expressions that, due to the verb they include, can only have low degrees of metaphoricity. Keywords: Distributional Semantics, Metaphoricity, Corpora 1. Introduction Metaphor has been defined as the exploitation of a norm in language: words have one or few frequent contexts of use that are considered normal (norms), and some unfre- quent contexts (exploitations) among which are metaphors, that are used to express new insights with a rhetorical effect (Hanks, 2004). Such an effect can have different degrees of strength, which correspond to the degrees of metaphoric- ity of the metaphor: the stronger the rhetorical effect, the higher the degree of metaphoricity. The usefulness of considering metaphoricity when inves- tigating and modelling metaphors has been stressed both from the theoretical point of view (Dunn, 2010; Hanks, 2004; Hanks, 2006; Nunberg, 1987) and in the field on NLP (Dunn, 2014; Dunn, 2013; Hovy et al., 2013; Mohler et al., 2015). (Shutova, 2015) in her vast review on metaphor pro- cessing systems suggests that real-world NLP applications should be concerned with the identification of metaphorical expressions with high degree of metaphoricity, i.e. those expressions that need to be interpreted differently from lit- eral language and therefore processed with specific tools. At the opposite, they should not address low-metaphoricity expressions, since their meanings are already present in dic- tionaries and can be interpreted using standard word sense disambiguation techniques. From this point of view, the degree of metaphoricity is a key element to distinguish between metaphors that require specific tools to be pro- cessed and those that do not. In spite of this, the author notes that the majority of current systems in NLP still ig- nore metaphoricity, assuming instead a binary distinction between literal …

Content truncated to 500 words