Filings in text format instead of htm

[rozeappletree]

Some 10-Q filings are in text format. The parser does not handle them very well.

For example, look at AT&T from 2003.

• filing detail: https://www.sec.gov/Archives/edgar/data/5907/0000005907-03-000252-index.htm
• 10-Q: https://www.sec.gov/Archives/edgar/data/5907/000000590703000252/thirdq10q2003.txt

You may use T1 as ticker symbol for sec-api.io in the demo streamlit app.

[Elijas]

Good catch! It's essential to bring up these edge cases.

Brainstorm idea 1
The text format still contains tags like <TABLE></TABLE> and separates out section titles with 1. CAPITALIZATION So in principle it should be possible to build a separate pipeline of Parsing Steps to parse text documents of this kind - but it's not clear how standard these text features (like <TABLE> tags in text files) are.

Brainstorm idea 2
Another way - if there's not a lot of those documents, especially if they have a different structure/features, some LLM pipeline could be used to parse those documents (long paragraphs could be con...densed to save the token count).

Do you maybe have any estimates of how widespread these text files are? As the vast majority of recent and upcoming 10-Q reports have a similar HTML format, we'll first focus on these, but let's keep a note to consider getting to the text format reports as well 💪

[rozeappletree]

I skimmed through this paper (https://arxiv.org/pdf/2104.11783.pdf) and found out there are text file format also. So, just found one example and tested it with sec-parser.

Look at:

2.2 File Type Classification

The SEC 10-Q filing typically has two types of formats – HTML
and text. Most filings in text format are antiquated, and more recent
filings from 2003 onwards are predominantly in HTML format.

[Blantranderozari]

Please check this GitHub project pipeline-sec-filings. Is this the one that you want?

[KakkoAki]

Thanks for sharing! We are monitoring them and need to conduct further tests as we're uncertain about the long-term reliability (adding more financial document types), scalability etc. So it needs some further evaluation and testing how we could use it. @Elijas