GitLocker: The Coding Marketplace

Description:

politicalml 0.4.0

API wrapper for the Political ML API
About Political ML
Political ML is a set of API that can perform multiple NLP related tasks. Currently three things are supported:

extracting entities
categorising texts
extracting meaningful content from a webpage or html

These tasks currently work best for Dutch texts. Political ML is developed by Wepublic. It only has a private API available, that's not meant for public use.
How to use
These clients are available: NerClient, CategoriseClient, ArticleExtractionClient. Below are some examples:
from political_ml import NerClient

client = NerClient(endpoint, token)
texts = [
{
"id": 1,
"text": "Mark Rutte eet wel bitterballen."
},
{
"id": 2,
"text": "Jesse Klaver eet geen bitterballen."
}
]
entities = client.ner(texts)

from political_ml import CategoriseClient

client = CategoriseClient(endpoint, token)
texts = [
{
"id": 1,
"text": "In Assendelft (gemeente Zaanstad, provincie Noord-Holland) is bij legkippen op een kleinschalige houderij vogelgriep (H5) vastgesteld. Het gaat waarschijnlijk om een hoogpathogene variant van de vogelgriep. Om verspreiding van het virus te voorkomen worden de circa 140 legkippen en 50 loopeenden van de besmette locatie geruimd. De ruiming wordt uitgevoerd door de Nederlandse Voedsel- en Warenautoriteit (NVWA)."
},
]
categories = client.categorise(texts)

from political_ml import ArticleExtractionClient

client = ArticleExtractionClient(endpoint, token)
meaningful_data = client.by_url("https://news.website/article/1")
other_meaningful_data = client.by_html("<html><body><h1>this is a title</h1><p>this is body text</p></body></html>")

For development
Requirements

Python 3.8+
Make

Setup for development

Create a virtual env: python -m virtualenv venv
Install development dependencies: python -m pip install -e ".[dev]"
Run make unit-test to run unit tests or run tox to run unit tests for all support python versions.
If you have an instance of the API's available, you can run integration tests with make integration-test.

New PRs are opened against develop.
Publishing a new version

Bump version numbers in meta.py
Publish a new release and create a git tag equal to the version number set in step 1.

A Github Action workflow takes care of building and publishing to PyPi.
Contact / maintainers
Jonathan (stakeholderintel@wepublic.nl) is the maintainer of this package.