Shared Latent Space of Font Shapes and Their Noisy Impressions

TL;DR

This paper introduces a method to embed fonts and their subjective impressions into a shared latent space, effectively capturing meaningful correlations despite noisy and irrelevant impression data.

Contribution

It proposes a novel approach using DeepSets to automatically enhance shape-relevant words and suppress irrelevant ones in the shared latent space.

Findings

Shared latent space accurately models font-impression correlations

Method effectively filters out noise from subjective impression words

Experimental results demonstrate improved correlation representation

Abstract

Styles of typefaces or fonts are often associated with specific impressions, such as heavy, contemporary, or elegant. This indicates that there are certain correlations between font shapes and their impressions. To understand the correlations, this paper realizes a shared latent space where a font and its impressions are embedded nearby. The difficulty is that the impression words attached to a font are often very noisy. This is because impression words are very subjective and diverse. More importantly, some impression words have no direct relevance to the font shapes and will disturb the realization of the shared latent space. We, therefore, use DeepSets for enhancing shape-relevant words and suppressing shape irrelevant words automatically while training the shared latent space. Quantitative and qualitative experimental results with a large-scale font-impression dataset demonstrate that the shared latent space by the proposed method describes the correlation appropriately, especially for the shape-relevant impression words.