[Feature] SEC - Management Discussion & Analysis Sections

[deeleeramone]

Beta Feature

1. Why?:

   • Extends the SEC functionality and brings in data that is not readily found programatically.

   • Key company disclosures are made within this text and it is not part of the structured financial statements.

2. What?:

   • New endpoint:
     • obb.equity.fundamental.management_discussion_analysis()
   • Parameters:
     • symbol - required
     • calendar year - optional
     • calendar period - optional
     • include_tables - optional (false)
     • raw_html - optional (false)
       • Use this if/when extraction fails, or you simply want the complete HTML filing.
     • strategy - optional ('trafilatura', 'inscriptis')
       • Defines the text extraction library to use. When "trafilatura" fails, "inscriptis" takes a pass.
       • Use this when table extraction was not successful via "trafilatura". One may do a better job than the other on a particular filing - example: MSFT
   • Returns:
     • Dictionary with the text string in the "content" key.
       • Text has markdown formatting for "**<h2><h3>** " and tables.

3. Impact:

   • Adds new dependencies:
     • https://trafilatura.readthedocs.io/en/latest/quickstart.html#with-python
     • https://github.com/weblyzard/inscriptis

4. Testing Done:

   • Various - WIP
   • Generally, issues will be related to "include_tables=True". Definitely open to help on broad solutions for huge variations in content.

5. Reviewer Notes:

   • You will need to run dev_install.py and then rebuild the static assets to run the function.

Example:

Example with tables:

Example Bad Tables (can potentially be saved post-request, information is discernible with some assembly required):

Successful Format Variation:

Embedded Images:

Example Failed Tables (MSFT) with Trafilatura:

Example "Fixed" by Inscriptis:

Some table headers, or individual rows, may be out of alignment. In those cases, they can be easily fixed with manual intervention. Neither method is perfect for generating valid markdown tables, it will get most of the way there, most of the time.