EXPRESS: Metaphor-Enabled Marketplace Sentiment Analysis

Abstract

Textual data requires an analytical tradeoff between breadth and depth. Automated approaches locate patterns across large swaths of data points but sacrifice qualitative insight because they are not well equipped to deal with context-determined ways to express meaning like figurative language. To strengthen the power of Automated Text Analysis (ATA), researchers seek hybrid methodologies where computer-augmented analysis is combined with sociocultural researcher insights based on qualitative textual interpretation. This article demonstrates a new method, that the authors term Metaphor-Enabled Marketplace Sentiment Analysis (MEMSA). Building on existing ATA methodologies linking word lists to sentiments, MEMSA adds metaphors which associate words or phrases across domains. Using MEMSA, researchers can leverage the sentiment potential of these located metaphors and scale insights to the level of big textual data by employing a dictionary approach enhanced by one unique useful linguistic property of metaphors: their predictable structure in text (something is something else). This article shows that metaphors add associative detail to sentiments revealing the targets and sources of sentiments that underlie the associations. Understanding nuanced market sentiments allows marketers to identify sentiment-based trends embedded in market discourse toward better formulating, targeting, positioning, and communicating value propositions for products and services.