juliushuizing / flint-precondition-labelling

Automatically extract preconditions for FLINT frames from Dutch sources of norms.

FLINT-Frame-Service

Automatically extract FLINT ACT frames from Dutch sources of norms.

Usage

Configuring your environment

• add a .env file to the root directory with the following keys:

OPENAI_SECRET_KEY=<YOUR_SECRET_KEY_HERE>

Getting started

• install and activate python 3.9.0*
• install poetry if new to poetry: https://python-poetry.org/docs/#installation
• (poetry env use /full/path/to/python)
• poetry config virtualenvs.in-project true
• poetry install
• Go to project settings of your editor and set the created virtual environment as project interpreter
• Important: this tool uses python 3.8 or higher because the transformers and the matplotlib package only support python 3.8 or higher. Running FlintFiller-srl with a lower python version results in an error.

Using the Act Frame Extractor Microservice

To start running the microservice for extracting act frames from user inputs, run:

flask --app flint_frame_service run

And request a response like:

curl -X POST http://127.0.0.1:5000/extract-act-frame \
-H "Content-Type: application/json" \
-d '{"article_string": "personal data shall be processed lawfully."}'

Note that by default, the server mimics processing time by sleeping for 5 - 10 seconds before returning a response. If you want disable this behavior, change the key of your request type to "debug":

curl -X POST http://127.0.0.1:5000/extract-act-frame \
-H "Content-Type: application/json" \
-d '{"debug": "personal data shall be processed lawfully."}'

Report:

• view only

flint-precondition-labelling's Issues

discuss first 100 labels together

identify distinct precondition categories

send mail to Meike & tom to clarify precondition categories

write script that prepares training data using preconditions only