This application connects to the US SEC archives EDGAR, and parses beneficial ownership information into a graph database. In this app, we leverage two layers of machine learning text enrichment models:
- NER (Named Entity Recognition)
- NRE (Nural Relationship Extraction)
The final product is mapped into a graph database.
Start up app with:
docker-compose build
docker-compose up
See below on Poetry dependency manager installation
poetry shell
bash start.sh
Poetry is a dependency managmnet tool for python. For setup, follow the setup on the main project repo. It is a bit of a hassle to set up, but the benefits are large for projects like this that depend on so many inter-dependent tools with multiple development and testing enviorments.
Once installed all dependencies can be installed with
poetry install
To enter into the virtual enviorment run
poetry shell
Currently we have to peg huggingface's transformers library at 3.5.1 due to conflicts with the modeles in OpenNRE
To utilize newer transformers like LUKE we will need to update to 4.0 library