WCXB: Web Content Extraction Benchmark with 2,008 Pages Across 7 Types

WCXB is the largest open benchmark for evaluating web content extraction, boilerplate removal, and main content detection. It provides 2,008 human-reviewed web pages spanning 7 page types and 1,613 domains, with ground truth annotations, HTML source files, and baseline results from 14 extraction systems. The dataset was created by murrough-foley and last updated on 2026-04 04.

Use Cases

• Benchmarking web content extraction systems based on the provided ground truth annotations and baseline results
• Training models for boilerplate removal based on the diverse HTML source files and page types
• Evaluating main content detection algorithms across non-news page types mentioned in the description
• Researching the performance of extraction tools on a large, domain-diverse collection of web pages

Strengths

• Contains 2,008 human-reviewed web pages, providing a substantial evaluation corpus
• Spans 7 distinct page types and 1,613 domains, offering diversity beyond typical news-only benchmarks
• Includes ground truth annotations, HTML source files, and baseline results from 14 extraction systems

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Last updated 2026-04-04 06:27:59; freshness should be verified

Provenance

Source

huggingface

Collection Method

Human-reviewed web pages with ground truth annotations.

Time Range

null

Freshness

Last updated 2026-04-04 06:27:59.

Geography

null