[Paper Review] Prospects for in-depth story understanding by computer

Erik T. Mueller|ArXiv.org|Mar 1, 2000

Topic Modeling31 references22 citations

TL;DR

This paper advocates for a renewed focus on in-depth story understanding in AI, proposing a multi-agent architecture to address long-standing challenges in narrative comprehension. It outlines key obstacles, solutions using interacting understanding agents, and identifies tools and resources to advance the field after progress in easier NLP tasks.

ABSTRACT

While much research on the hard problem of in-depth story understanding by computer was performed starting in the 1970s, interest shifted in the 1990s to information extraction and word sense disambiguation. Now that a degree of success has been achieved on these easier problems, I propose it is time to return to in-depth story understanding. In this paper I examine the shift away from story understanding, discuss some of the major problems in building a story understanding system, present some possible solutions involving a set of interacting understanding agents, and provide pointers to useful tools and resources for building story understanding systems.

Motivation & Objective

• To re-engage research in in-depth story understanding after a shift toward simpler NLP tasks in the 1990s.
• To identify core challenges in building systems capable of deep narrative comprehension.
• To propose a multi-agent architecture as a framework for integrating different aspects of story understanding.
• To provide practical tools and resources to support the development of story understanding systems.
• To position story understanding as a viable and important next frontier in natural language processing.

Proposed method

• Propose a system composed of interacting understanding agents, each responsible for a distinct aspect of story comprehension.
• Leverage advances in information extraction and word sense disambiguation as foundational components.
• Integrate modular agents to handle events, characters, goals, and causal relationships in narratives.
• Use existing NLP tools and resources to support agent functionality and system integration.
• Design the architecture to allow dynamic coordination among agents for holistic story interpretation.
• Draw on prior research to inform agent design and ensure compatibility with current NLP pipelines.

Experimental results

Research questions

• RQ1 What are the key challenges that have hindered progress in in-depth story understanding by computers?
• RQ2 How can a modular, agent-based architecture effectively model complex narrative structures?
• RQ3 What role do recent advances in information extraction and word sense disambiguation play in enabling deeper story understanding?
• RQ4 Which existing tools and resources can be leveraged to build scalable story understanding systems?
• RQ5 How can multiple understanding agents coordinate to achieve a unified interpretation of a story?

Key findings

• The field has shifted focus from story understanding to easier NLP tasks like information extraction and word sense disambiguation since the 1990s.
• A multi-agent system offers a viable framework for addressing the complexity of in-depth story understanding.
• Progress in simpler NLP tasks now provides a solid foundation for returning to story understanding.
• The proposed architecture enables modular, coordinated processing of narrative elements such as events, characters, and goals.
• A range of existing tools and resources are available to support the implementation of story understanding agents.
• The paper provides a roadmap for future research by identifying key components and integration strategies.