Hugging Face's logo

Datasets:
PeytonT
/
paper_graph

Modalities:
Tabular
Text
Formats:
parquet
Size:
10M - 100M
Tags:
datasets
graph
scientific-papers
arxiv
retrieval
embeddings
Libraries:
Datasets
Dask
Polars
Dataset card Data Studio Files
xet
 Community
1
paper_graph / README.md
PeytonT's picture
PeytonT
Update README.md
b942fac verified
preview code
|
raw
history blame contribute delete
4.52 kB
metadata 
pretty_name: Paper Universe Graph
viewer: true
tags:
  - datasets
  - graph
  - scientific-papers
  - arxiv
  - retrieval
  - embeddings
size_categories:
  - 10M<n<100M
configs:
  - config_name: paper_nodes
    default: true
    data_files:
      - split: train
        path: paper_nodes/*.parquet
  - config_name: paper_category_edges
    default: false
    data_files:
      - split: train
        path: paper_category_edges/*.parquet
  - config_name: paper_knn
    default: false
    data_files:
      - split: train
        path: paper_knn/*.parquet
  - config_name: category_nodes
    default: false
    data_files:
      - split: train
        path: category_nodes/*.parquet
  - config_name: category_knn
    default: false
    data_files:
      - split: train
        path: category_knn/*.parquet
  - config_name: topic_nodes
    default: false
    data_files:
      - split: train
        path: topic_nodes/*.parquet
  - config_name: paper_topic_edges
    default: false
    data_files:
      - split: train
        path: paper_topic_edges/*.parquet
  - config_name: year_nodes
    default: false
    data_files:
      - split: train
        path: year_nodes/*.parquet
  - config_name: paper_year_edges
    default: false
    data_files:
      - split: train
        path: paper_year_edges/*.parquet
  - config_name: paper_embeddings
    default: false
    data_files:
      - split: train
        path: paper_embeddings/*.parquet
  - config_name: paper_fulltext_embeddings
    default: false
    data_files:
      - split: train
        path: paper_fulltext_embeddings/*.parquet

Generated with Research Library:

https://github.com/peytontolbert/Research_Library

Paper Universe Graph Dataset

Parquet-first export of the paper-universe graph already built under the Repository Library.

This dataset preserves:

  • paper nodes with metadata references and 3D coordinates
  • paper/category/year/topic graph layers
  • optional paper-to-paper and category-to-category similarity edges
  • metadata and full-text paper embedding splits

Coverage

  • papers: 1000000
  • categories: 156
  • years: 19
  • topics: 1483957
  • embedding dimension: 384
  • full-text embeddings included: true

Configs

  • default viewer/config: paper_nodes
  • paper_nodes: 1000000 rows
  • paper_category_edges: 1744492 rows
  • paper_knn: 20000062 rows
  • category_nodes: 156 rows
  • category_knn: 1248 rows
  • topic_nodes: 1483957 rows
  • paper_topic_edges: 3000000 rows
  • year_nodes: 19 rows
  • paper_year_edges: 1000000 rows
  • paper_embeddings: 1000000 rows
  • paper_fulltext_embeddings: 1000000 rows

Loading 

from datasets import load_dataset

paper_nodes = load_dataset("PeytonT/paper_graph", "paper_nodes")
paper_category_edges = load_dataset("PeytonT/paper_graph", "paper_category_edges")
paper_knn = load_dataset("PeytonT/paper_graph", "paper_knn")
category_nodes = load_dataset("PeytonT/paper_graph", "category_nodes")
category_knn = load_dataset("PeytonT/paper_graph", "category_knn")
topic_nodes = load_dataset("PeytonT/paper_graph", "topic_nodes")
paper_topic_edges = load_dataset("PeytonT/paper_graph", "paper_topic_edges")
year_nodes = load_dataset("PeytonT/paper_graph", "year_nodes")
paper_year_edges = load_dataset("PeytonT/paper_graph", "paper_year_edges")
paper_embeddings = load_dataset("PeytonT/paper_graph", "paper_embeddings")
paper_fulltext_embeddings = load_dataset("PeytonT/paper_graph", "paper_fulltext_embeddings")

Notes

  • paper_nodes stores metadata references and coordinates, not the full paper body.
  • The original full text remains in the source paper dataset referenced by the manifest.
  • paper_embeddings is the metadata/title+abstract embedding split.
  • paper_fulltext_embeddings is the aggregated full-body embedding split when available.

Visualization Assets

The export includes the local paper-universe visualizations and viewer payload when present:

  • manifest.json
  • progress.json
  • render_manifest.json
  • viewer_manifest.json
  • universe_3d.png
  • universe_3d_detailed.png
  • nodes_3d_sample.html
  • universe_3d_hover.html
  • interactive/categories.json
  • interactive/manifest.json
  • interactive/papers_200000.json
  • interactive/papers_50000.json
  • interactive/years.json

Universe 3D Overview

Paper universe 3D overview

Universe 3D Detailed View

Paper universe 3D detailed view