Notice

Dataset Card generated by Kimi K2.6

Dataset Card for philpapers_260508

Dataset Summary

This dataset contains ~1.13 million unique works crawled from PhilPapers, a comprehensive index and bibliography of philosophy literature. Each row represents a single bibliographic work with its metadata, abstract (where available), and PhilPapers category assignments.

The dataset preserves titles, authors, abstracts, publication details, PhilPapers categories, and crawl metadata for each entry.

---

Dataset Details

Dataset Sources

• Source Website: https://philpapers.org

---

Uses

Direct Use

Suitable for:

• Academic NLP research and training (philosophy-domain language models, citation analysis, abstract summarisation)
• Bibliometric and scientometric research in philosophy and related humanities
• Category classification and topic modelling across philosophical sub-disciplines
• Multilingual humanities text analysis (130 languages represented)
• Information retrieval and semantic search over philosophical literature

---

Dataset Structure

Schema

Column	Type	Description
rec_id	string	PhilPapers record identifier (e.g. PLAFAR)
title	string	Work title
authors	list<string>	Author names
abstract	string	Abstract text (may be empty)
year	string	Publication year (raw string; some non-standard values present)
publication	string	Publication venue or publisher
doi	string	DOI (empty in this snapshot)
language	string	ISO language code (e.g. en, de, fr)
categories	list<string>	PhilPapers category labels assigned to this work
extra_versions	int64	Count of additional versions listed on PhilPapers
detail_url	string	Direct URL to the PhilPapers entry page
crawled_at	string	ISO-8601 timestamp of crawl
crawl_error	string	Error message if crawl failed (empty for all rows in this snapshot)

Statistics

Metric	Value
Documents	1,128,673
Columns	13
Languages	130
Abstracts with text	~556K (49.3%)
Mean abstract length (non-empty)	~950 characters
Median abstract length (non-empty)	~851 characters
Mean authors per work	~1.31
Mean categories per work	~1.80
Works with extra versions	72,633 (6.4%)

Top languages by document count

Language	Documents	Approx. share
English (en)	977,316	~86.6%
German (de)	34,419	~3.0%
French (fr)	26,924	~2.4%
Spanish (es)	23,715	~2.1%
Italian (it)	12,567	~1.1%
Portuguese (pt)	9,189	~0.8%

Top categories by document count

Category	Documents
Biomedical Ethics	40,351
Ethics	33,368
Classics	25,765
Philosophy of Education	22,319
Business Ethics	17,650
Social and Political Philosophy	14,451
Political Theory	13,865
Philosophy of Religion	13,088
Logic and Philosophy of Logic	13,058
Medical Ethics	11,331

Top authors by occurrence count

Author	Occurrences
Terence Rajivan Edward	727
Barry Smith	712
Xiaoyang Yu	638
George Sarton	542
Immanuel Kant	463

Coverage completeness

Field	Non-empty rate
rec_id, title, detail_url, crawled_at, language	100%
authors	100% (list present; may be empty list)
categories	100% (list present; may be empty list)
publication	~95.9%
year	~100% (raw value present; ~95.6% parseable as 4-digit year)
abstract	~49.3% (non-empty string)
doi	~0%
crawl_error	~0%

---

Dataset Creation

Curation Rationale

PhilPapers is the largest curated index of philosophical literature. A structured crawl makes this corpus available in a machine-readable tabular format for bibliometrics, NLP, and digital humanities research.

Source Data

Data Collection and Processing

• Tool: Scrapy
• Scope: All works discoverable via PhilPapers' category browse pages.
• No content filtering was applied; all discoverable bibliographic entries are included.
• Deduplication: The raw crawl contained duplicate rows because the same work can appear under multiple category pages. The dataset was deduplicated by rec_id so that each row represents one unique work.

Who are the source data producers?

Authors, editors, and publishers of philosophical literature. PhilPapers is a non-profit initiative that aggregates metadata and, where available, abstracts from publishers, repositories, and direct user submissions.

---

Bias, Risks, and Limitations

Content Risks

• The dataset includes abstracts and titles that may discuss sensitive philosophical topics including ethics of violence, suicide, eugenics, discrimination, and other contested subjects.
• Some works are historical texts that reflect attitudes now considered offensive or discriminatory.

Structural Limitations

• Abstract coverage: Only ~49% of works have non-empty abstracts.
• Year field: Stored as raw strings with occasional non-standard values (e.g. date ranges, prefixes, or malformed data). Only ~96% are readily parseable as 4-digit years.
• DOI field: Empty in this snapshot; DOIs were not successfully extracted during the crawl.
• Language coverage: While 130 languages are present, the corpus is heavily skewed toward English.
• Category granularity: Categories range from broad areas (e.g. "Ethics") to very narrow philosopher-specific tags (e.g. "Russell: Axioms"), leading to a long-tail distribution.
• The dataset represents a snapshot in time; PhilPapers content is continuously updated.

Personal and Sensitive Information

• No systematic PII detection or redaction was performed.
• Author names and publication details are present as bibliographic metadata.
• The dataset should be treated as containing personal information of living and deceased scholars.

Licence and Redistribution

No licence was identified for the source PhilPapers content. This dataset is therefore marked as license: unknown.

The dataset consists of scraped bibliographic metadata and abstracts. The dataset publisher does not claim ownership of the underlying text and does not grant additional rights beyond any rights users may already have under applicable law.

Users are responsible for determining whether their intended use is lawful, including with respect to copyright, privacy, data protection, and platform terms.