Linguistic creativity

Name: Geert Brône’s & Tony Veale

Ever since linguistics emerged as a cognitive science in the 1960′s, linguistic creativity has been a key concept in the discipline. In Chomskyan linguistics, the notion of a Universal Grammar is rooted in a particular property of language, i.e. “that it provides the means for expressing indefinitely many thoughts and for reacting appropriately in an indefinite range of new situations” (Chomsky 1965: 6). On this definition, creativity resides in the ability to produce a (potentially) infinite number of novel structures/utterances on the basis of existing ones (grammatical competence), the output of which is in essence predictable. Obviously, however, linguistic creativity is not restricted to the level of predictable grammatical creativity. Rather, creativity is often associated with inexplicability (Humboldt 1836) or deliberate nonconformity. So, at a different level, creativity entails aspects of novelty or surprise that go beyond the level of innovation generated by purely productive creativity in the Chomskyan sense, and in a sense seems to even go against the purely rule-based generativist definition. At the level of full creativity (Bergen & Binsted 2004), novelty is associated with an element of unexpectedness, emergence and insightfulness.

The art of linguistic creativity – and humorous creativity in particular -, however, lies not in finding new truths to express in language (if, indeed, there are any), but in finding new and more resonant ways to express truths that are already known or implicitly accepted. Creative expression thus requires that we adopt a new and revealing perspective on a familiar idea, one that may prompt an audience to conceptually re-represent this idea to draw out new meaning or new relevance to a situation. Linguistic and computational models of creativity require a knowledge representation that is as semantically agile and accommodating as the creative speakers who use it. In this talk, we discuss:

  1. contemporary models of linguistic creativity and humor – mainly rooted in cognitive linguistics – that aim to uncover the principles and mechanisms of flexible representation (including Giora’s ‘optimal innovation hypothesis’, Fauconnier & Turner’s ‘conceptual integration theory’ and Coulson’s ‘space structuring model’).
  2. the computational requirements of such a flexible knowledge representation. We present an initial computational realization of this representation (called Talking Points) that is derived from large text corpora.