Mapping News Narratives Using LLMs and Narrative-Structured Text Embeddings

Jan Elfes

2024-09-10

Abstract:Given the profound impact of narratives across various societal levels, from personal identities to international politics, it is crucial to understand their distribution and development over time. This is particularly important in online spaces. On the Web, narratives can spread rapidly and intensify societal divides and conflicts. While many qualitative approaches exist, quantifying narratives remains a significant challenge. Computational narrative analysis lacks frameworks that are both comprehensive and generalizable. To address this gap, we introduce a numerical narrative representation grounded in structuralist linguistic theory. Chiefly, Greimas' Actantial Model represents a narrative through a constellation of six functional character roles. These so-called actants are genre-agnostic, making the model highly generalizable. We extract the actants using an open-source LLM and integrate them into a Narrative-Structured Text Embedding that captures both the semantics and narrative structure of a text. We demonstrate the analytical insights of the method on the example of 5000 full-text news articles from Al Jazeera and The Washington Post on the Israel-Palestine conflict. Our method successfully distinguishes articles that cover the same topics but differ in narrative structure.

Computation and Language

What problem does this paper attempt to address?

### The Problem the Paper Attempts to Solve

The paper aims to address the lack of a comprehensive and generalizable framework in computational narrative analysis. Specifically, the authors propose a numerical narrative representation method based on structuralist linguistic theory to capture the semantics and narrative structure of texts. Through this method, the paper hopes to analyze and compare narrative trends across different topics and sources.

### Background and Motivation

1. **Importance of Narratives**:
- Narratives have a profound impact on various social levels, including personal identity and international politics.
- In online spaces, narratives can quickly spread and exacerbate social divisions and conflicts.

2. **Shortcomings of Existing Research**:
- Despite the existence of many qualitative methods, quantifying narratives remains a significant challenge.
- Computational narrative analysis lacks a comprehensive and generalizable framework.

### Solution

1. **Numerical Narrative Representation**:
- Based on Greimas' actantial model, which represents narratives through six functional roles (subject, object, sender, receiver, helper, opponent).
- Using open-source large language models (LLMs) to extract these actants and integrate them into narrative-structured text embeddings to capture the semantics and narrative structure of texts.

2. **Application Example**:
- The authors demonstrate the analytical capability of this method on 5000 news articles from Al Jazeera and The Washington Post about the Israel-Palestine conflict.
- The method can distinguish articles that report on the same topic but have different narrative structures.

### Main Contributions

1. **Providing a Comprehensive and Generalizable Narrative Representation Method**:
- The method is based on structuralist linguistic theory and is suitable for narrative analysis across different subjects and sources.

2. **Empirically Validating the Method's Effectiveness**:
- By analyzing a large number of news articles, the method identifies 18 different narrative trends, revealing editorial narrative differences between different news sources.

### Conclusion

By introducing a new numerical narrative representation method, the paper fills a gap in the field of computational narrative analysis. This method not only captures the semantic content of texts but also reflects their narrative structure, providing a powerful tool for analyzing narrative trends across topics and sources.