Constructions and Frames - Volume 8, Issue 2, 2016

Conceptual Metaphor Theory (CMT) aims to represent the conceptual structure of metaphors rather than the structure of metaphoric language. The theory does not explain which aspects of metaphoric language evoke which conceptual structures, for example. However, other theories within cognitive linguistics may be better suited to this task. These theories, once integrated, should make building a unified model of both the conceptual and linguistic aspects of metaphor possible. First, constructional approaches to syntax provide an explanation of how particular constructional slots are associated with different functions in evoking metaphor. Cognitive Grammar is especially effective in this regard. Second, Frame Semantics helps explain how the words or phrases that fill the relevant constructional slots evoke the source and target domains of metaphor. Though these theories do not yet integrate seamlessly, their combination already offers explanatory benefits, such as allowing generalizations across metaphoric and non-metaphoric language, and identifying the words that play a role in evoking metaphors, for example.

This paper describes an innovative formalization of Conceptual Metaphor Theory and its implementation in a structured metaphor repository. Central to metaphor analysis is the development of an internal structure of frames and relations between frames, based on an Embodied Construction Grammar framework, which then informs the structure of metaphors and relationships between metaphors. The hierarchical nature of metaphors and frames is made explicit, such that inferential information originating in embodied conceptual primitives is inherited throughout the network. The present analysis takes a data-driven approach, where lexical differences in linguistic expressions attested in naturally-occurring discourse lead to a continued refinement and expansion of our analyses.

Public discourse on highly charged, complex social and political issues is extensive, with millions of sentences available for analysis. It is also rife with metaphors that manifest vast numbers of novel metaphoric expressions. More and more, to understand such issues, to see who is saying what and why, we require big data and statistically-based analysis of such corpora. However, statistically-based data processing alone cannot do all the work. The MetaNet (MN) project has developed an analysis method that formalizes existing insights about the conceptual metaphors underlying linguistic expressions into a computationally tractable mechanism for automatically discovering new metaphoric expressions in texts. The ontology used for this computational method is organized in terms of metaphor cascades, i.e. pre-existing packages of hierarchically organized primary and general metaphors that occur together. The current paper describes the architecture of metaphor-to-metaphor relations built into this system. MN’s methodology represents a proof of concept for a novel way of performing metaphor analysis. It does so by applying the method to one particular domain of social interest, namely the gun debate in American political discourse. Though well aware that such an approach cannot replace a thorough cognitive, sociological, and political analysis, this paper offers examples that show how a cascade theory of metaphor and grammar helps automated data analysis in many ways.

This paper demonstrates the fruitful application of the formalization of Conceptual Metaphor Theory, combined with metaphor constructions and computational tools to a large-scale, corpus-based approach to the study of metaphor expressions. As the case study of poverty metaphor expressions illustrates, the representation of individual metaphors and frames as parts of larger conceptual networks facilitates analyses that capture both local details and larger patterns of metaphor use. Significantly, the data suggest that the two most frequently used source domain networks in poverty metaphor expressions each support different types of inferences about poverty, its effects, and possible ways to reduce or end it.

This paper presents MetaNet’s automatic metaphor detection system that applies theoretical principles from construction grammar, frame semantics, and recent developments in conceptual metaphor theory, including the theory of cascades (Lakoff 2014). The system has achieved relative success in identifying metaphorical expressions for a range of target domains from large corpora and holds promise as a useful tool for corpus-based study of metaphor. The detection system relies on MetaNet’s conceptual network of frames and metaphors as a computational resource for its functionality, and improves automatically as the representations stored in the network are built up. In addition, because of its theoretically principled design the system’s level of accuracy at identifying metaphorical expressions provides feedback to linguists about the accuracy of the frame and metaphor analyses in the network.