This repo implements a document pre-processing pipeline for SEC filings. Currently, the pipeline is capable of extracting narrative text from user-specified sections in 10-K, 10-Q, and S-1 filings.
-
Using
pyenvto manage virtualenv's is recommended -
Create a virtualenv to work in and activate it, e.g. for one named
sec-filings:pyenv virtualenv 3.8.15 sec-filings
pyenv activate sec-filings -
Run
make install -
Start a local jupyter notebook server with
make run-jupyter
OR
just start the fast-API locally withmake run-web-app
You can run this Colab notebook to see how pipeline-section.ipynb extracts the narrative text sections from an SEC Filing and defines an API.
To retrieve narrative text section(s) from an iXBRL S-1, 10-K, or 10-Q document (or amended version S-1/A, 10-K/A, or 10-Q/A), post the document to the /section API. You can try this out by downloading the sample documents using make dl-test-artifacts. Then, from
the sample-docs folder, run:
curl -X 'POST' \
'http://localhost:8000/sec-filings/v0.2.1/section' \
-H 'accept: application/json' \
-H 'Content-Type: multipart/form-data' \
-F '[email protected]' \
-F section=RISK_FACTORS | jq -C . | less -R
Note that additional -F section parameters may be included in the curl request to fetch
multiple sections at once. Valid sections for 10-Ks,
10-Qs, and S-1s
are available on the SEC website. You can also reference
this file
for a list of valid section parameters, e.g. RISK_FACTORS OR MANAGEMENT_DISCUSSION.
You'll get back a response that looks like the following. Piping through jq and less
formats/colors the outputs and lets your scroll through the results.
{
"RISK_FACTORS": [
{
"text": "You should carefully consider the risks described in this section. Our future performance is subject to risks and uncertainties that could have a material adverse effect on our business, results of operations, and financial condition and the trading price of our common stock. We may be subject to other risks and uncertainties not presently known to us. In addition, please see our note about forward-looking statements included in the MD&A.",
"type": "NarrativeText"
},
{
"text": "Our revenue is subject to volatility in metal prices, which could negatively affect our results of operations or cash flow.",
"type": "NarrativeText"
},
{
"text": "Market prices for gold, silver, copper, nickel, and other metals may fluctuate widely over time and are affected by numerous factors beyond our control. These factors include metal supply and demand, industrial and jewelry fabrication, investment demand, central banking actions, inflation expectations, currency values, interest rates, forward sales by metal producers, and political, trade, economic, or banking conditions.",
"type": "NarrativeText"
},
...
]
}
You can also pass in custom section regex patterns using the section_regex parameter. For
example, you can run the following command to request the risk factors section:
curl -X 'POST' \
'http://localhost:8000/sec-filings/v0.2.1/section' \
-H 'accept: application/json' \
-H 'Content-Type: multipart/form-data' \
-F '[email protected]' \
-F 'section_regex=risk factors' | jq -C . | less -R
The result will be:
{
"REGEX_0": [
{
"text": "You should carefully consider the risks described in this section. Our future performance is subject to risks and uncertainties that could have a material adverse effect on our business, results of operations, and financial condition and the trading price of our common stock. We may be subject to other risks and uncertainties not presently known to us. In addition, please see our note about forward-looking statements included in the MD&A.",
"type": "NarrativeText"
},
{
"text": "Our revenue is subject to volatility in metal prices, which could negatively affect our results of operations or cash flow.",
"type": "NarrativeText"
},
{
"text": "Market prices for gold, silver, copper, nickel, and other metals may fluctuate widely over time and are affected by numerous factors beyond our control. These factors include metal supply and demand, industrial and jewelry fabrication, investment demand, central banking actions, inflation expectations, currency values, interest rates, forward sales by metal producers, and political, trade, economic, or banking conditions.",
"type": "NarrativeText"
},
...
]
}
As with the section parameter, you can request multiple regexes by passing in multiple values
for the section_regex parameter. The requested pattern will be treated as a raw string.
You can also use special regex characters in your pattern, as shown in the example below:
curl -X 'POST' \
'http://localhost:8000/sec-filings/v0.2.1/section' \
-H 'accept: application/json' \
-H 'Content-Type: multipart/form-data' \
-F '[email protected]' \
-F "section_regex=^(\S+\W?)+$"
You can always replace the header -H 'accept: application/json' with -H 'accept: text/csv' depending on the format you want to fetch from the API as follows:
curl -X 'POST' \
'http://localhost:8000/sec-filings/v0.2.1/section' \
-H 'accept: text/csv' \
-H 'Content-Type: multipart/form-data' \
-F '[email protected]' \
-F section=RISK_FACTORS | jq -C . | less -R
The result will be:
"section,element_type,text\r\nRISK_FACTORS,NarrativeText,\"You should carefully consider the risks described in this section. Our future performance is subject to risks and uncertainties that could have a material adverse effect on our business, results of operations, and financial condition and the trading price of our common stock. We may be subject to other risks and uncertainties not presently known to us. In addition, please see our note about forward-looking statements included in the MD&A.\"\r\nRISK_FACTORS,NarrativeText,\"Our revenue is subject to volatility in metal prices, which could negatively affect our results of operations or cash flow.\"\r\nRISK_FACTORS,NarrativeText,\"Market prices for gold, silver, copper, nickel, and other metals may fluctuate widely over time and are affected by numerous factors beyond our control. These factors include metal supply and demand, industrial and jewelry fabrication, investment demand, central banking actions, inflation expectations, currency values, interest rates, forward sales by metal producers, and political, trade, economic, or banking conditions.\"\r\n
In addition, you can add the form -F 'output_schema=labelstudio' if you want an output to be compatible with labelstudio as follows:
curl -X 'POST' \
'http://localhost:8000/sec-filings/v0.2.1/section' \
-H 'accept: application/json' \
-H 'Content-Type: multipart/form-data' \
-F '[email protected]' \
-F 'output_schema=labelstudio' \
-F section=RISK_FACTORS | jq -C . | less -R
The result will be:
{
"RISK_FACTORS": [
{
"data": {
"text": "You should carefully consider the risks described in this section. Our future performance is subject to risks and uncertainties that could have a material adverse effect on our business, results of operations, and financial condition and the trading price of our common stock. We may be subject to other risks and uncertainties not presently known to us. In addition, please see our note about forward-looking statements included in the MD&A.",
"ref_id": "7a912bb639b547404be4ceaf5d9083a9"
}
},
{
"data": {
"text": "Our revenue is subject to volatility in metal prices, which could negatively affect our results of operations or cash flow.",
"ref_id": "d4cc8e0e0c2b68ef69282c5250b721c9"
}
},
...
]
}
You can use some of the functions provided in prepline_sec_filings.fetch to directly view or manipulate the filings available from the SEC's EDGAR API.
For example, get_filing(cik, accession_number, your_organization_name, your_email) will return the text of the filing with accession number accession_number for the organization with CIK number cik.
your_organization_name and your_email should be your information.
The parameters your_organization_name and your_email are passed along to Edgar's API to identify the caller and are required by Edgar.
Alternatively, the parameters may be omitted if the environment variables SEC_API_ORGANIZATION and SEC_API_EMAIL are defined.
Helper functions are also provided for cases where the CIK and/or accession numbers are not known. For example,
get_form_by_ticker('mmm', '10-K', your_organization_name, your_email) returns the text of the latest 10-K filing from 3M,
and open_form_by_ticker('mmm', '10-K', your_organization_name, your_email) opens the SEC index page for the same filing in a web browser.
The python module section.py contains the FASTApi code needed to serve the API. It's created with make generate-api, which derives the API from the notebook pipeline-section.ipynb.
You can generate the FastAPI APIs from all pipeline-notebooks/ by running make generate-api.
It is not necessary to run Docker in a local development environment, however a Dockerfile and
make targets of docker-build, docker-start-api, and docker-start-jupyter are provided for convenience.
You can also launch a Jupyter instance to try out the notebooks with 
See our security policy for information on how to report security vulnerabilities.
| Section | Description |
|---|---|
| Company Website | Unstructured.io product and company info |
| EDGAR API | Documentation for the SEC |
| 10-K Filings | Detailed documentation on 10-K filings |
| 10-Q Filings | Detailed documentation on 10-Q filings |
| S-1 Filings | Detailed documentation on S-1 filings |
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent