The emotional arcs of stories are dominated by six basic shapes (1606.07772v3)

Abstract: Advances in computing power, natural language processing, and digitization of text now make it possible to study a culture's evolution through its texts using a "big data" lens. Our ability to communicate relies in part upon a shared emotional experience, with stories often following distinct emotional trajectories and forming patterns that are meaningful to us. Here, by classifying the emotional arcs for a filtered subset of 1,327 stories from Project Gutenberg's fiction collection, we find a set of six core emotional arcs which form the essential building blocks of complex emotional trajectories. We strengthen our findings by separately applying Matrix decomposition, supervised learning, and unsupervised learning. For each of these six core emotional arcs, we examine the closest characteristic stories in publication today and find that particular emotional arcs enjoy greater success, as measured by downloads.

• The paper identifies six fundamental emotional arcs by applying sentiment analysis, SVD, and clustering techniques to 1,327 works of fiction.
• The study demonstrates that these arcs correlate with narrative success, suggesting a universal appeal in story structures.
• The paper’s methodology integrates unsupervised learning and statistical analysis, offering a replicable framework for computational narratology.

Analysis of "The emotional arcs of stories are dominated by six basic shapes"

This paper presents an empirical analysis of narrative structures based on a large corpus of texts from Project Gutenberg. Of particular interest is the identification of six basic emotional arcs that serve as the foundational structural elements in storytelling. Using computational techniques, the authors have explored the tonal trajectories that give shape to narratives and used these structures to discern patterns that might correlate with story types and authorial success.

Methodology and Techniques

The authors employ a combination of natural language processing (sentiment analysis), statistical methods, and machine learning techniques to categorize the emotional tones of 1,327 works of fiction. Key computational techniques include:

1. Sentiment Analysis: By quantifying the emotional content of text, sentiment analysis provides a temporal sentiment score across individual story arcs.
2. Singular Value Decomposition (SVD): SVD is utilized for dimensionality reduction, revealing principal components of variation within the emotional trajectories of stories. This method helps in identifying the leading modes of emotional change in the narrative corpus.
3. Hierarchical Clustering: Books are sorted into clusters based on the similarity of their emotional arcs, indicating common story arcs.
4. Self-Organizing Maps (SOM): These are used for the unsupervised identification of recurring patterns in the temporal data, showing the self-organized convergence toward the identified arcs.

Findings

The analysis reveals six fundamental emotional arcs considered canonical:

• Rise ("Rags to Riches"): A consistent upward trajectory.
• Fall ("Tragedy/Riches to Rags"): A decline across the arc.
• Fall-Rise ("Man in a Hole"): An initial fall followed by a rise.
• Rise-Fall ("Icarus"): Initial improvement ending in a decline.
• Rise-Fall-Rise ("Cinderella"): Up, down, and up again.
• Fall-Rise-Fall ("Oedipus"): Down, recovery, and down.

The robust identification of these arcs across different analytical methods bolsters their claim of universality. Moreover, the success of stories bearing these arcs, measured in terms of download counts, indicates a potential link between the emotional arcs and narrative success.

Implications and Future Work

From the perspective of computational narratology and literary analysis, the implications are significant. These findings suggest a fundamental set of emotional trajectories that are perhaps universally appealing, lending themselves to cross-cultural examinations of storytelling. Practically, this paper holds implications for industries invested in narrative craft—such as publishing and entertainment—providing a foundation for the synthetic generation or adaptation of compelling narratives.

Theoretically, it raises questions about the cognitive and psychological underpinnings of narrative preference and the archetypical nature of the identified arcs. Future research could explore the sociocultural variables that modulate these arcs and investigate their impact over different time periods and genres. Additionally, deeper integration with character network analysis and plot extraction could yield comprehensive insights into narrative structure and evolution.

In conclusion, this research contributes significantly to understanding narrative structures and their universal appeal. The convergence of computational methods in story analysis presents a powerful platform for exploring narratives, potentially impacting both theoretical frameworks in the humanities and practical applications in storytelling industries.