Time-Track Narrative Graph

TTNG is my PhD research project on making the structure of unfolding news narratives explicit, inspectable, and editable instead of leaving it implicit in long-form text.

What I built

• A directed graph model representing how stories develop across time and themes.
• A graph-to-text pipeline using language models to generate controlled news material with known narrative structure.
• TTNG Constructor, which combines prompt-based generation, live search, and drag-and-drop narrative motifs on a time-track grid.

Evaluation

I designed and ran a controlled study with 50 participants. Visual views helped participants answer analysis questions 15-51% faster than text-only reading, while highly abstract views revealed an accuracy trade-off for complex comprehension tasks.