About

Propp is a Python-based Natural Language Processing (NLP) framework developed for the analysis of narrative texts, with a focus on long-form literary fiction.

Designed for computational literary studies, Propp bridges Natural Language Processing and narrative theory to enable character-centric, ontology-based, and corpus-scale analysis of prose narratives.

The framework is designed for research on characters, narrative structure, and large-scale literary corpora, while remaining accessible to scholars with limited programming experience.

Objectives

Bridging narratological theory and contemporary research in natural language processing.

Narratology Context

Vladimir Propp

Computational Humanities

Design Philosophy

Modularity
Open Source
Informed by narratology theory

French Fiction and Beyond

While the starting point of the project is long form French fiction narratives (novels) we aim at exploring other types of nnarratives.

Dream narratives, altered state of consciousness narratives, personal narratives, myths and legends.

Project Lineage

2014 - Java BookNLP
- A Bayesian Mixed Effects Model of Literary Character (Bamman et al., ACL 2014)
2019 - BookNLP
- An annotated dataset of literary entities (Bamman et al., NAACL 2019)
2020 - Multilingual BookNLP
2021 - BookNLP-fr
- BookNLP-fr, the French Versant of BookNLP. A Tailored Pipeline for 19th and 20th Century French Literature (Mélanie et al., JCLS 2024)
2025 - Propp

Modules

Propp framework combines deep learning and rule-based techniques for:

Token Embeddings

Contextualized embeddings from pretrained transformer-based models (e.g., google/mt5-xl, almanach/camembert-large).
Mention-Spans Detection

Identifying spans in the text that refer to named or nominal entities, as well as referential pronouns.
Mention Type Prediction

Classifying detected mentions as persons, locations, organizations, geopolitical entities.
Coreference Resolution

Linking mentions that refer to the same underlying entity across the narrative.
Characters Detection

Consolidating coreferent mentions into coherent character entities.
Character-Related Attributes Extraction

Inferring gender, social role, narrative function, and other descriptors from context.
Character Representation

Producing structured profiles for characters (e.g., via embeddings), ready for downstream analysis or visualization.

Funding

This project was funded in part by the PRAIRIE–PSAI (Paris Artificial Intelligence Research Institute – Paris School of Artificial Intelligence), reference ANR-22-IACL-0008.

The project also received support from the Major Research Program of PSL Research University, "CultureLab", launched by PSL Research University and implemented by the French National Research Agency (ANR), reference ANR-10-IDEX-0001.

The project additionally benefited from the support of the Fondation de l’ENS, through the Be Ys Chair, dedicated to the support of innovative research projects in artificial intelligence.