sec-parsers

SEC Parsers

Parses non-standardized SEC filings into structured xml. Use cases include LLMs, NLP, and textual analysis. Average parse-time for a 100 page document is 0.4 seconds. Package is a WIP, and is updated frequently.

Supported filing types are 10-K, 10-Q, 8-K, S-1, 20-F. More will be added soon, or you can write your own! How to write a Custom Parser in 5 minutes

sec-parsers is maintained by John Friedman, and is under the MIT License. If you use sec-parsers for a project, please let me know! Feedback

URGENT: Advice on how to name functions used by users is needed. I don't want to deprecate function names in the future. Link

Notice download_sec_filing is being deprecated.

Installation

pip install sec-parsers # base package
pip install sec-parsers['all'] # installs all extras
pip install sec-parsers['downloaders'] # installs downloaders extras
pip install sec-parsers['visualizers'] # installs visualizers extras

Quickstart

Load package

from sec_parsers import Filing

Downloading html file (new)

from sec_downloaders import SEC_Downloader

downloader = SEC_Downloader()
downloader.set_headers("John Doe", "[e-mail]")
download = downloader.download(url)
filing = Filing(download)

Downloading html file (old)

from sec_parsers download_sec_filing
html = download_sec_filing('https://www.sec.gov/Archives/edgar/data/1318605/000162828024002390/tsla-20231231.htm')
filing = Filing(html)

Parsing

filing.parse() # parses filing
filing.visualize() # opens filing in webbrowser with highlighted section headers
filing.find_sections_from_title(title) # finds section by title, e.g. 'item 1a'
filing.find_sections_from_text(text) # finds sections which contains your text
filing.get_tree(node) # if no argument specified returns xml tree, if node specified, returns that nodes tree
filing.get_title_tree() # returns xml tree using titles instead of tags. More descriptive than get_tree.
filing.get_subsections_from_section() # get children of a section
filing.get_nested_subsections_from_section() # get descendants of a section
filing.set_filing_type(type) # e.g. 'S-1'. Use when automatic detection fails
filing.save_xml(file_name,encoding='utf-8')
filing.save_csv(file_name,encoding='ascii')

Additional Resources:

• quickstart
• How to write a Custom Parser in 5 minutes
• Archive of Parsed XMLs / CSVs - Last updated 7/24/24.
• example parsed filing
• example parsed filing exported to csv.

Features:

• lots of filing types
• export to xml, csv, with option to convert to ASCII
• visualization

Feature Requests:

Request a Feature

• company metadata (sharif) - will add to downloader
• filing metadata (sharif) - waiting for SEC Downloaders first release
• Export to dta (Denis)
• DEF 14A, DEFM14A (Denis)
• Export to markdown (Astarag)
• Better parsing_string handling. Opened an issue. (sharif)

SEC Downloader

Not released yet, different repo.

• Download by company name, ticker, etc
• Download all 10-Ks, etc
• Rate limit handling
• asynchronous downloads

Statistics

• Speed: On average, 10-K filings parse in 0.25 seconds. There were 7,118 10-K annual reports filed in 2023, so to parse all 10-Ks from 2023 should take about half an hour.
• Improving speed is currently not a priority. If you need more speed, let me know. I think I can increase parsing speed to ~ .01 seconds per 10-K.

Other packages useful for SEC filings

• https://github.com/dgunning/edgartools

Updates

Towards Version 1:

• Most/All SEC text filings supported
• Few errors
• xml

Might be done along the way:

• Faster parsing, probably using streaming approach, and combining modules together.
• Introduction section parsing
• Signatures section parsing
• Better visualization interface (e.g. like pdfviewer for sections)

Beyond Version 1:

To improve the package beyond V1 it looks like I need compute and storage. Not sure how to get that. Working on it.

Metadata

• Clustering similar section titles using ML (e.g. seasonality headers)
• Adding tags to individual sections using small LLMs (e.g. tag for mentions supply chains, energy, etc)

Other

• Table parsing
• Image OCR
• Parsing non-html filings

Current Priority list:

• look at code duplication w.r.t to style detectors, e.g. all caps and emphasis. may want to combine into one detector

• yep this is a priority. have to handle e.g. Introduction and Segment Overview as same rule. Bit difficult. Will think over.

• better function names - need to decide terminology soon.
• consider adding table of contents, forward looking information, etc

• forward looking information, DOCUMENTS INCORPORATED BY REFERENCE, TABLE OF CONTENTS - go with a bunch,

• fix layering issue - e.g. top div hides sections
• make trees nicer
• add more filing types
• fix all caps and emphasis issue
• clean text
• Better historical conversion: handle if PART I appears multiple times as header, e.g. logic here item 1 continued.