Modeling Human Mental States with an Entity-based Narrative Graph
This work addresses narrative understanding for applications in natural language processing, but appears incremental as it builds on existing methods with task-adaptive pre-training and symbolic inference.
The paper tackled the problem of understanding narrative text by capturing characters' motivations, goals, and mental states, proposing an Entity-based Narrative Graph (ENG) to model these internal states and evaluating it on tasks like predicting character mental states and desire fulfillment.
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.