Modeling Human Mental States with an Entity-based Narrative Graph

TL;DR

This paper introduces an Entity-based Narrative Graph (ENG) that models characters' mental states and interactions in stories, improving narrative understanding by capturing internal states and dependencies.

Contribution

It presents a novel graph-based approach for modeling character mental states, incorporating explicit entity interactions and task-adaptive training methods.

Findings

Effective in predicting character mental states

Improves desire fulfillment understanding

Provides qualitative insights into narrative modeling

Abstract

Understanding narrative text requires capturing characters' motivations, goals, and mental states. This paper proposes an Entity-based Narrative Graph (ENG) to model the internal-states of characters in a story. We explicitly model entities, their interactions and the context in which they appear, and learn rich representations for them. We experiment with different task-adaptive pre-training objectives, in-domain training, and symbolic inference to capture dependencies between different decisions in the output space. We evaluate our model on two narrative understanding tasks: predicting character mental states, and desire fulfillment, and conduct a qualitative analysis.