PeS2o Filtered: Openly Licensed Academic Papers from S2ORC

A version of the peS2o dataset restricted to openly licensed articles, derived from the S2ORC corpus of academic papers. The dataset was created by the common-pile organization and was last updated on June 6, 2025. It contains papers converted to a structured format using Grobid, with filtering applied for quality and language.

Use Cases

• Training language models for scientific domains based on the corpus of academic papers.
• Information extraction research based on the structured format of the papers.
• Analyzing trends in open-access scientific literature based on the openly licensed subset.
• Developing text classification models for academic genres based on the filtered paper content.

Strengths

• Derived from the large-scale S2ORC corpus, which suggests a substantial base collection.
• Includes quality filtering for paper length, metadata, language (English), and OCR errors.
• Restricted to openly licensed articles, which facilitates legal reuse.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

S2ORC (Semantic Scholar Open Research Corpus)

Collection Method

Derived from S2ORC using Grobid for structured conversion, followed by heuristic- and model-based filtering.

Freshness

Last updated 2025-06-06 03:58:18; freshness should be verified.