Railway of Emotions: Visualizing Per-Character Sentiment in Narrative Text

Abstract

The readers of long-form narratives find it challenging to track character emotions, differentiate emotional states, and stay engaged. Current sentiment visualization systems either aggregate emotions across characters or use disconnected diagrams requiring visual switching. We introduce SentiViz, a railway-themed system that employs a unified metaphor to represent real-time, per-character emotions during narrative reading, where characters are trains, emotions are trackside scenes, and plot structure is a train map with intersecting tracks. It uses a layered update strategy ambient lighting changes instantly for emotional awareness, while scene content transitions gradually to minimize distraction. A pilot study with ten participants reading two stories (with and without SentiViz) showed significant improvements in identifying main emotions (p = 0.016) and understanding emotional shifts (p = 0.031), with effect sizes of +1.40 and +1.50 on a 5-point scale. The system also scored significantly higher in aesthetics (p = 0.008) and recommendation likelihood (p = 0.047). Exploratory analysis indicated greater benefits for low-confidence readers. This work demonstrates that unified metaphors with real-time, per-character sentiment visualization can effectively support emotional interpretation during narrative reading, particularly for ESL learners and similar groups.