Metaphor clusters in discourse

Abstract

The phenomenon of clustering, where speakers or writers suddenly produce multiple metaphors, is widespread and intriguing. This paper presents an innovative visualisation methodology for identifying and exploring metaphor clusters, comparing it to existing methods that use cumulative frequency graphs and Poisson curve fitting, and addressing issues that arise from these. Identification of clusters from the visualisation is shown to be reliable and practical, while also offering in-depth exploration across a range of discourse parameters. Conversations aimed at conciliation between a perpetrator of violence and a victim (total 160 minutes) are analysed for clusters and their discourse functions. All techniques show clusters at two distinct time scales, of around one minute and of several seconds. Clusters in conciliation talk account for about 42 per cent of the total metaphors, and cover about 30 per cent of the discourse. Discourse work carried out in clusters includes explanation of a speaker’s perspective to the Other, appropriation of metaphors originally used by the Other, and exploration of alternative, negative, scenarios that had been possible choices for the speaker but had been rejected. The finding that metaphor clusters are sites of intensive work relating to the central discourse purpose supports cluster exploration as a heuristic tool for discourse analysis.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 56 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.