Tracing metaphors in time through self-distance in vector spaces

TL;DR

This paper proposes a method to trace metaphorical language evolution over time by analyzing changes in cosine similarity of word vectors in diachronic Italian corpora, linking shifts to metaphor emergence.

Contribution

It introduces a novel approach using self-distance in vector spaces to detect the emergence of metaphors across different time periods.

Findings

Cosine similarity drops correlate with documented metaphor emergence.

Method successfully identifies metaphorical shifts in historical language data.

Provides a new quantitative tool for diachronic metaphor analysis.

Abstract

From a diachronic corpus of Italian, we build consecutive vector spaces in time and use them to compare a term's cosine similarity to itself in different time spans. We assume that a drop in similarity might be related to the emergence of a metaphorical sense at a given time. Similarity-based observations are matched to the actual year when a figurative meaning was documented in a reference dictionary and through manual inspection of corpus occurrences.