The SEC publishes financial novels. This splits them by chapter.

What it does

EDGAR-CRAWLER is a Python toolkit that scrapes SEC EDGAR for 10-K, 10-Q, and 8-K filings and parses the unstructured HTML into structured JSON keyed to standard item sections. It filters downloads by year, quarter, filing type, and company identifier, returning metadata and the text of specific items—item_1A for risk factors, item_7 for management discussion, and so on—ready for downstream NLP or quantitative analysis. The toolkit handles the full pipeline from raw download to cleaned output, including retrieval of the EDGAR index files used to locate reports.

The interesting bit

The parser maps each filing type to its specific regulatory item taxonomy, so a 10-Q’s part_1_item_2 or an 8-K’s item_5.02 land in predictable JSON keys. For older 10-Q documents where item-level extraction fails, it gracefully falls back to part-level buckets rather than silently dropping data.

Key highlights

• Supports 10-K, 10-Q, and 8-K filings with item-specific JSON keys.
• Filters downloads by year, quarter, filing type, and company identifier (CIK or ticker).
• Includes full metadata: filing date, period of report, SIC code, and direct SEC links.
• Backed by an academic paper presented at WWW 2025 and a related HuggingFace dataset (EDGAR-CORPUS).
• Falls back to Part-level text for older 10-Q filings when item detection isn’t possible.

Caveats

• The toolkit is driven by a config.json file; the README shows no interactive CLI or programmatic API surface.
• Older 10-Q filings may collapse to part_1 and part_2 blobs instead of individual items.

Verdict

Worth a look if you are building financial NLP models or need machine-readable SEC text without writing your own HTML scraper. Skip it if you are after real-time market data or already have a Bloomberg terminal.