Prediction of sentence "nature" in a French political sentence

This model aims at predicting the nature of a sentence in a French political sentence. The predictions fall in three categories:

problem: the sentence describes a problem (usually to be tackled by the speaker), for example il y a dans ce pays une fracture (J. Chirac)
solution: the sentences describes a solution (typically part of a political programme), for example: J’ai supprimé les droits de succession parce que je crois au travail et parce que je crois à la famille. (N. Sarkozy)
other: the sentence does not belong to any of these categories, for example: vive la République, vive la France

This model was trained using AutoNLP based on sentences extracted from a mix of political tweets and speeches.

Model Trained Using AutoNLP

Problem type: Multi-class Classification
Model ID: 23105051
CO2 Emissions (in grams): 1.06099358268878

Validation Metrics

Loss: 0.6050735712051392
Accuracy: 0.8097826086956522
Macro F1: 0.7713543865034599
Micro F1: 0.8097826086956522
Weighted F1: 0.8065488494385247
Macro Precision: 0.7861074705111403
Micro Precision: 0.8097826086956522
Weighted Precision: 0.806470454156932
Macro Recall: 0.7599656456873758
Micro Recall: 0.8097826086956522
Weighted Recall: 0.8097826086956522

Usage

You can use cURL to access this model:

$ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "Il y a dans ce pays une fracture"}' https://api-inference.huggingface.co/models/mazancourt/politics-sentence-classifier

Or Python API:

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("mazancourt/autonlp-politics-sentence-classifier-23105051", use_auth_token=True)

tokenizer = AutoTokenizer.from_pretrained("mazancourt/politics-sentence-classifier", use_auth_token=True)

inputs = tokenizer("Il y a dans ce pays une fracture", return_tensors="pt")

outputs = model(**inputs)

# Category can be "problem", "solution" or "other"
category = outputs[0]["label"]
score = outputs[0]["score"]