Cognitive Linguistics: Current Applications and Future Perspectives
The various contributions: foundations, applications and future perspectives
Virtual reality as a new experience
The study of the computational applications to CL is a comparatively new area of enquiry which poses a challenge both for CL, which needs to be tested in terms of its adaptability to computational requirements, and for computer science, which needs to assess its own ability to cope with the descriptive and explanatory richness of an approach to language that explicitly ties together body, mind, thinking and language.
In this context of mutual challenges, John Barnden’s chapter, entitled “Artificial intelligence, figurative language and cognitive linguistics”, discusses the relationship between CL and Artificial Intelligence. The chapter is focused on Barnden’s well-known ATT-Meta project, which is both an explanatory approach and a computer program capable of performing reasoning tasks involved in metaphorical utterances. The overall project is related to but at the same time challenges some of the central assumptions in Lakoff’s (1993) conceptual metaphor theory. However, as Barnden himself remarks, ATT-Meta is in some respects fairly close to Grady’s theory of primary metaphor (Grady 2005). Thus, rather than a mapping from machines to minds, ATT-Meta prefers to postulate transfer rules from ‘physical operation’ to ‘mental operation’. Also significantly, the ATT-Meta approach questions the usefulness of the notion of domain in computational implementations because of the high degree of arbitrariness involved in determining domain boundaries. This leads Barnden– together with other AI theorists– to disregard domain mappings and see metaphor as a matter of transferring from a pretense to reality. This move has implications for the traditional distinction between metaphor and metonymy in terms of domain-external versus domain-internal mappings. For Barnden, metaphor and metonymy are merely vague notions playing a heuristic role, while the real phenomena lie at a deeper level involving conceptual distance, similarity, contiguity, and connections between source and target, an approach which thus ties in with Panther’s criticism of the unreliability of the notion of ‘domain’.
Unlike Barnden’s contribution, the chapter by Tony Veale, entitled “Computability as a test on linguistics theories”, focuses more on computational assessment than on mutual feedback between CL and AI. The chapter explores the explanatory power of three CL models of language processing– Attardo’s general theory of verbal humor (Attardo et al. 2002), Fauconnier and Turner’s blending theory (Fauconnier and Turner 2002) and Croft’s Radical Construction Grammar (Croft 2001)– in terms of several computational criteria, among them specificity, efficiency, and tractability. Computational specificity is a test of explicitness whereby every step and every data element in a process must be clearly specified. Efficiency is measured against the ability of a theory to avoid unnecessary proliferation of representations, constraints, and processes. Tractability reflects the ability to run sufficiently explicit computer specifications in a “reasonable” amount of time. The exploration allows Veale to show how the computational perspective brings to the fore a number of strengths and weaknesses in each approach. For example, while the multiple-input account provided by blending theory fares poorly in terms of efficiency, the general theory of verbal humor is more constrained in this respect. However, Attardo’s theory falls short of providing a sufficiently explanatory account of processing. The chapter ends by discussing, together with a number of structure-mapping issues, the computational feasibility of Radical Construction Grammar (RCG). All in all, Veale’s chapter suggests that computationalism can not only assess the explanatory power of a linguistic theory but also allow researchers to gain a clearer understanding of linguistic phenomena by making sure that terms and their corresponding notions have a precise theoretical meaning from the perspective of computational sufficiency.
