Wikipedia PT Categories: 2,873 Portuguese Articles for Clustering

Wikipedia PT Categories is a Portuguese clustering evaluation dataset containing 2,873 articles from pt.wikipedia.org, each labeled with one of 15 broad topic categories. The dataset was created by tardellirs and serves as the source for the WikipediaPTCategoriesClusteringP2P task in the MTEB(por) benchmark. It was last updated on 2026-06-08.

Use Cases

• Benchmarking Portuguese text clustering models based on paragraph-level article text.
• Evaluating unsupervised topic discovery methods based on Wikipedia article content.
• Training or fine-tuning Portuguese text classifiers using the 15 provided topic labels.
• Studying the structure and topical coverage of Portuguese Wikipedia as a corpus.

Strengths

• Contains 2,873 individual data points (articles).
• Provides 15 distinct, labeled topic categories for evaluation.
• Serves as a standardized benchmark task within the MTEB(por) leaderboard.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known but total file size and specific data formats are unknown.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Portuguese Wikipedia (pt.wikipedia.org)

Collection Method

Built directly from Portuguese Wikipedia via the MediaWiki API.

Freshness

Last updated 2026-06-08 02:11:26; freshness should be verified.