Is Sentiment Banana-Shaped? Exploring the Geometry and Portability of Sentiment Concept Vectors

TL;DR

This paper evaluates the portability and assumptions of Concept Vector Projections (CVP) in sentiment analysis across diverse domains, languages, and affective dimensions, highlighting its effectiveness and areas for improvement.

Contribution

It systematically assesses CVP's transferability across genres, periods, and languages, and examines the linearity assumption underlying the method.

Findings

CVP trained on one corpus transfers well to others with minimal performance loss.

CVP effectively captures generalizable sentiment patterns across domains and languages.

Linearity assumption of CVP is approximate, indicating room for methodological improvements.

Abstract

Use cases of sentiment analysis in the humanities often require contextualized, continuous scores. Concept Vector Projections (CVP) offer a recent solution: by modeling sentiment as a direction in embedding space, they produce continuous, multilingual scores that align closely with human judgments. Yet the method's portability across domains and underlying assumptions remain underexplored. We evaluate CVP across genres, historical periods, languages, and affective dimensions, finding that concept vectors trained on one corpus transfer well to others with minimal performance loss. To understand the patterns of generalization, we further examine the linearity assumption underlying CVP. Our findings suggest that while CVP is a portable approach that effectively captures generalizable patterns, its linearity assumption is approximate, pointing to potential for further development. Code available at: github.com/lauritswl/representation-transfer