Semantic Interaction for Narrative Map Sensemaking: An Insight-based Evaluation

arXiv cs.CL / 4/1/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper investigates semantic interaction (SI) for narrative map sensemaking, aiming to let analysts directly manipulate visualizations to steer AI-driven narrative extraction using their own cognitive processes.
A user study with 33 participants compares a timeline baseline, a basic narrative map, and an SI-enabled interactive narrative map, finding map-based prototypes produce more insights than timelines.
The SI-enabled condition achieves statistical significance and the basic narrative map trends similarly, with large effect sizes (d > 0.8) between map conditions suggesting the study may have been underpowered.
Qualitative findings distinguish two SI usage styles—corrective and additive—that allow analysts to apply quality judgments and organizational structure to the extracted narratives.
SI users attain comparable exploration breadth with less parameter manipulation, indicating SI can be an alternative route for “model refinement” beyond tuning parameters.

Abstract

Semantic interaction (SI) enables analysts to incorporate their cognitive processes into AI models through direct manipulation of visualizations. While SI frameworks for narrative extraction have been proposed, empirical evaluations of their effectiveness remain limited. This paper presents a user study that evaluates SI for narrative map sensemaking, involving 33 participants under three conditions: a timeline baseline, a basic narrative map, and an interactive narrative map with SI capabilities. The results show that the map-based prototypes yielded more insights than the timeline baseline, with the SI-enabled condition reaching statistical significance and the basic map condition trending in the same direction. The SI-enabled condition showed the highest mean performance; differences between the map conditions were not statistically significant but showed large effect sizes (d > 0.8), suggesting that the study was underpowered to detect them. Qualitative analysis identified two distinct SI approaches-corrective and additive-that enable analysts to impose quality judgments and organizational structure on extracted narratives. We also find that SI users achieved comparable exploration breadth with less parameter manipulation, suggesting that SI serves as an alternative pathway for model refinement. This work provides empirical evidence that map-based representations outperform timelines for narrative sensemaking, along with qualitative insights into how analysts use SI for narrative refinement.