Comparable-titles graph
This is the SageChimp neighbor graph: for each catalog title, the other titles that share genre and theme overlap, ranked the same way MCP comparable titles are ranked. It is bibliographic metadata only — no review text, no descriptions, no user data.
A neighbor shares a genre when the source has genres. Theme overlap is required only when both titles have themes. Neighbors are ranked by genre and theme overlap. Lists that never clear a quality floor stay empty or show at most two titles. Targets that already appear on many lists are penalized.
- Catalog books
- 6,047
- License
- CC-BY-4.0
- Generated
- live, streamed
The JSONL dumps cover the full catalog. Edges are scored from a compact in-memory index and flushed in chunks so the web process never holds the whole graph as one string.
Download
- comparable-titles.jsonl — one neighbor edge per line
- books.jsonl — catalog nodes (title, authors, ISBN, genres, themes, URLs)
- comparable-titles.json — counts, method, file list
Mirrors
- Hugging Face mirror after the dataset is published on an org account.
- Zenodo: DOI after the first snapshot publish
- Wikidata batch: wikidata.qs
Load in Python
from datasets import load_dataset
edges = load_dataset("json", data_files="https://www.sagechimp.com/datasets/comparable-titles.jsonl", split="train")
books = load_dataset("json", data_files="https://www.sagechimp.com/datasets/books.jsonl", split="train")Cite as SageChimp comparable-titles graph, https://www.sagechimp.com. Prefer the Zenodo DOI once it exists.