Co-Reading: A Human-AI Co-Reader Media System for Poetry-Emotion-Color
Date:
Co-Reading is a multimodal system in which humans and AI share an interpretive environment for literary texts, translating poetry into color-based visuals as a third mode of reading.
Author
Iro Lim and Byungjun Kim
Abstract
This paper proposes Co-Reading, a multimodal system that lets humans and AI share an interpretive environment for literary texts by transforming poetry into color-based visuals. We argue that such a system constitutes a third mode of reading, one that combines the interpretive depth of close reading with the scalability of distant reading. The system works in three stages. First, emotion classification draws on the KPoEM dataset, comprising 7,662 annotated lines of Korean poetry labeled with 44 fine-grained emotions. Second, the KCoEM dataset maps those emotions onto color palettes grounded in Korean color psychology. Third, generative AI is integrated into a real-time web prototype that renders the resulting visualization dynamically as a reader moves through a poem. The project reframes datasets not merely as analytical instruments but as media that enable new sensory experiences and interpretive possibilities. Its contribution to digital humanities is threefold: it extends human-AI interaction beyond information exchange toward shared interpretation, it introduces an emotion-centered alternative to text-based visualization, and it positions datasets as environmental media that connect research infrastructure to cultural application.