Dataset Card for zyda_urls

This dataset provides the URLs and top-level domains associated with training records in Zyphra/Zyda. It is part of a collection of datasets curated to make exploring LLM training datasets more straightforward and accessible.

Dataset Details

Dataset Description

This dataset was created by downloading the source data, extracting URLs and top-level domains, and retaining only those record identifiers. In doing so, it allows researchers and practitioners to explore the contents of these training datasets without having to manage terabytes of raw text. You can explore the pipeline used to construct this dataset on GitHub.

• Curated by: Nick Hagar and Jack Bandy
• License: Same as source dataset

Dataset Sources

• Repository: Zyphra/Zyda

Uses

This dataset is intended to allow researchers and practitioners to analyze the contents of large LLM training datasets without having to wade through terabytes of unwieldy text data.

Direct Use

The main use case for these data is to explore the contents of LLM training datasets at scale. This might involve:

• Identifying the most-used websites
• Categorizing URLs to understand domain- or topic-level dataset composition
• Comparing URLs across datasets
• Digging into inclusion/exclusion patterns for a particular website

Out-of-Scope Use

This dataset is not intend to replicate or replace the source data, nor is it intended to enable large-scale scraping of the URLs listed. For source text, refer to the original dataset.

Dataset Structure

This dataset contains every record with a URL from the source dataset. It contains two columns:

• url: The raw URL associated with each record
• domain: The top-level domain for each URL, extracted with tldextract

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]