Color2Vec: Web-Based Modeling of Word-Color Association with Sociocultural Contexts

Abstract

Color design has long benefited from the statistical analysis of public taste and, more recently, from crowdsourcing to discover fresh and popular ideas. However, the current color dictionary is considerably restricted in terms of the scope of expressible design concepts and the control of target demographics. We propose a search-engine-based color palette generator inspired by Natural Language Processing algorithms that filter and cluster semantically related words. The post-evaluation reveals that our results not only faithfully realize the given keywords but are notable indicators of inter-group dynamics; the differential recognition of the other group's identity colors reflects the direction of historic, geographic, or cultural influence.