BRANCHES: A Dataset of Narrative Graphs from Visual Novels

BRANCHES (Branching Narrative Corpus for Human-Interactive Entertainment Systems) is a dataset of narrative structures extracted from publicly available visual novels developed with the Ren'Py engine. Each project is represented as a directed narrative graph, with accompanying dialogue, player choices, and structural metrics.

The dataset was introduced in the paper:

BRANCHES: A Dataset and Analysis of Narrative Graphs in Visual Novels
XXV Simpósio Brasileiro de Jogos e Entretenimento Digital (SBGames 2026)

---

Dataset Summary

Statistic	Value
Total projects	592
GitHub projects	183
itch.io projects	409
Total dialogue lines	~2.6M
Total narrative nodes	~49K
Total edges	~73K

---

Data Sources

Games were collected from two sources:

• GitHub — public repositories collected via the GitHub search API using Ren'Py-specific queries targeting .rpy file extensions and structural script patterns.
• itch.io — publicly available games discovered through genre and engine-related tags including visual-novel, renpy, otome, story-rich, romance, yuri, yaoi, anime, dating-sim, and narrative. Only freely downloadable games with a Ren'Py relevance score ≥ 0.4 were included.

---

Dataset Structure

BRANCHES/
├── summary.csv                 # Per-game metrics for all 592 projects
├── all_dialogue.csv            # Unified dialogue across all games
├── games.csv                   # itch.io metadata (ratings, tags, price, etc.)
├── coverage_report.json        # Pipeline parse coverage statistics
├── pipeline/
│   └── parser_multifile.py     # Full parsing pipeline
└── dataset/
    └── {game_id}/
        ├── dialogue.csv        # Dialogue and narration lines
        ├── choices.csv        # Player choice options and targets
        ├── nodes.csv          # Narrative labels with terminal node flags
        ├── edges.csv          # Typed directed edges
        └── metadata.json      # Per-game structural statistics

File descriptions

summary.csv — one row per game, with the following columns:

Column	Description
game_id	Unique identifier (extracted folder name)
source_folder	Original path on disk
num_labels	Number of narrative labels
num_dialogue_lines	Total dialogue and narration lines
num_choices	Number of player choices
num_nodes	Number of graph nodes
num_edges	Number of directed edges
num_conditional_edges	Edges arising from if/elif/else blocks
num_terminal_nodes	Nodes with no outgoing edges
max_path_length	Longest acyclic path in the narrative graph
avg_path_length	Mean path length across all entry-to-terminal paths

dialogue.csv (per game) — columns: id, node, character, text

choices.csv (per game) — columns: node, choice, target

nodes.csv (per game) — columns: node, is_terminal

edges.csv (per game) — columns: from, to, type, condition

Edge types: jump, call, choice, conditional

metadata.json (per game) — JSON with all structural statistics including terminal_node_ratio, num_conditional_edges, and path length metrics.

games.csv (itch.io subcorpus only) — columns: url, title, author, description, score, rating, rating_count, price, is_free, age_restricted, tags, platforms, download_pages

---

Parsing Pipeline

The parsing pipeline is included in pipeline/parser_multifile.py. It implements an indentation-aware static parser for Ren'Py .rpy script files, supporting:

• Character alias resolution (define x = Character("Name"))
• Multi-line string joining
• Indentation-aware menu and choice scope tracking
• Conditional branch extraction (if/elif/else)
• call vs jump edge type distinction
• Explicit terminal node detection via return statements

To reparse from source scripts:

pip install pandas
python pipeline/parser_multifile.py path/to/game/folder

Or to run on all games at once:

python pipeline/parser_multifile.py

---

Content Advisory

A subset of the dataset contains content intended for adult audiences. Among the itch.io subcorpus, 36 games (8.8%) were flagged as age-restricted based on mature content indicators on their game pages. These games are retained to preserve structural diversity. Researchers working with dialogue content should apply appropriate filtering using the age_restricted field in games.csv.

The GitHub subcorpus does not include equivalent content ratings. Researchers should exercise independent judgment when working with the full dataset.

---

Licensing and Copyright

The structural data in this dataset (graph topology, metrics, and metadata) is released under CC BY-NC 4.0.

The dialogue content (text extracted from visual novel scripts) is derived from works authored by independent developers and is provided for non-commercial research purposes only, consistent with fair use principles. Raw script files are not redistributed. Users are responsible for ensuring compliance with applicable copyright law in their jurisdiction.

Takedown requests

Game authors who wish to have their content removed from this dataset may contact us at [carbinatti@alunos.utfpr.edu.br]. Please include the game title, itch.io or GitHub URL, and the specific content you wish to have removed. Requests will be reviewed and processed within 30 days.

---

Citation

If you use BRANCHES in your research, please cite:

@inproceedings{branches2026,
  author    = {Carbinatti, Giovanni and Gibaut, Wandemberg},
  title     = {BRANCHES: A Dataset and Analysis of Narrative Graphs in Visual Novels},
  booktitle = {XXV Simpósio Brasileiro de Jogos e Entretenimento Digital (SBGames 2026)},
  year      = {2026},
  address   = {Goiânia, GO}
}

---

Acknowledgements

We thank the independent developers who made their visual novel projects publicly available on GitHub and itch.io, making this dataset possible.