Datasets:

konsman
/

physics-corpus

language:
  - en
license: cc-by-4.0
task_categories:
  - text-generation
  - text-retrieval
tags:
  - physics
  - arxiv
  - scientific-text
  - ontology
  - sections
configs:
  - config_name: quantum-physics
    data_files:
      - split: train
        path: quantum-physics/train-*.parquet

Physics Corpus — `konsman/physics-corpus`

arXiv physics papers exported from a PostgreSQL mirror of the Kaggle arXiv dataset, structured for ontology extraction and downstream NLP pipelines.

Configuration: `quantum-physics`

arXiv categories included: quant-ph, hep-th, gr-qc

Schema

Field	Type	Description
`paper_id`	string	`arxiv:<id>v<n>` — stable across pipeline runs
`arxiv_id`	string	Base arXiv ID without version suffix
`arxiv_version`	int32	Version number (for deduplication)
`doi`	string	DOI or empty string
`content_hash`	string	SHA-256 of `full_text` — for delta detection
`title`	string	Paper title
`abstract`	string	Paper abstract
`authors`	list[string]	Author names
`arxiv_categories`	list[string]	e.g. `["hep-th", "gr-qc"]`
`primary_category`	string	First listed category
`journal`	string	Journal reference or empty string
`keywords`	list[string]	Keywords (empty if unavailable)
`submission_date`	string	ISO date of first submission
`processed_date`	string	ISO date this record was exported
`full_text`	string	All section texts joined by `\n\n`
`sections`	list	Structured sections (see below)

Section fields

Field	Type	Description
`section_type`	string	`TITLE` / `ABSTRACT` / `BACKGROUND` / `METHOD` / `RESULTS` / `DISCUSSION` / `CONCLUSION` / `OTHER`
`section_title`	string	Raw heading text, e.g. `3. Experimental Setup`
`text`	string	Section body text

Record structure

Each row is one paper. sections is a list of objects — one per section:

{
  "paper_id": "arxiv:2401.00123v2",
  "arxiv_id": "2401.00123",
  "arxiv_version": 2,
  "title": "...",
  "abstract": "...",
  "authors": ["Author A", "Author B"],
  "arxiv_categories": ["quant-ph"],
  "primary_category": "quant-ph",
  "full_text": "Introduction text\n\nMethod text\n\n...",
  "sections": [
    {"section_type": "BACKGROUND", "section_title": "1. Introduction", "text": "..."},
    {"section_type": "METHOD",     "section_title": "2. Methods",      "text": "..."},
    {"section_type": "RESULTS",    "section_title": "3. Results",      "text": "..."}
  ]
}

Usage

from datasets import load_dataset

ds = load_dataset("konsman/physics-corpus", "quantum-physics", split="train")

for paper in ds:
    print(paper["paper_id"], paper["primary_category"])
    for sec in paper["sections"]:
        print(sec["section_type"], sec["section_title"], sec["text"][:80])

Source

Papers are sourced from the Kaggle arXiv dataset via a PostgreSQL mirror. Only papers with a JOIN match in the metadata table are included (≥ 96% of sections have a match).

License

Dataset card and structure: CC-BY-4.0. Paper texts are subject to the individual arXiv paper licenses.

Physics Corpus — konsman/physics-corpus

Configuration: quantum-physics