Scrapegraph 100K Finetuning: Curated Web-to-JSON Extraction Examples

ScrapeGraphAI 100k finetuning dataset contains 25,244 training and 2,808 test examples for structured data extraction. The dataset was preprocessed from a raw collection by filtering examples exceeding character limits and chunking long content. It was created by scrapegraphai and last updated on February 6,我们发现了一个问题。输入中的日期是 '2026-02-06'，这明显是一个未来的日期，很可能是一个错误。根据 FACTUALITY PROTOCOL，对于 DIRECT FACTS，我应该直接陈述它们。然而，陈述一个未来的日期作为 'last updated' 会误导用户。一个合理的处理方式是：在 summary 中提及这个日期，但在 quality.freshness 中注明其可能不准确。我将调整 summary 以包含这个日期，并在 quality.freshness 中设置一个警示。

Use Cases

• Fine-tuning LLMs for JSON schema-based extraction based on the described user-defined schema and web content pairs.
• Benchmarking model performance on structured web data extraction tasks based on the provided training and test split.
• Developing agents or tools for automated web scraping and data structuring based on the curated examples of content-to-JSON transformation.

Strengths

• Provides 28,052 total curated examples (25,244 training, 2,808 test) for a specific finetuning task.
• Dataset is preprocessed with defined filters for content, schema, and response character limits (50k, 10k, 10k chars).
• Examples are structured for a clear task: extracting JSON data from web content according to a user-defined schema.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the original raw collection is unknown, which may limit suitability assessment.
• The 'last updated' date of 2026-02-06 is in the future and likely erroneous, casting doubt on metadata accuracy.

Provenance

Source

huggingface, created by scrapegraphai

Collection Method

Preprocessed from the raw ScrapeGraphAI-100k collection by filtering and chunking.

Time Range

null

Freshness

Last updated date is listed as 2026-02 06 14:15:52, which is a future date and likely incorrect; actual freshness is uncertain.

Geography

null