metadata
license: cc-by-4.0
datasets:
- clarin-pl/poquad
language:
- pl
library_name: transformers
Model Card
Extractive Question-Answer model for polish. Extractive means, that the most relevant chunk of the text is returned as answer from the context for the given question.
Model Details
- Model name:
radlab/polish-roberta-large-v2-qa - Developed by: radlab.dev
- Shared by: radlab.dev
- Model type: QA
- Language(s) (NLP): PL
- License: CC-BY-4.0
- Finetuned from model: sdadas/polish-roberta-large-v2
- Maxiumum context size: 512 tokens
Model Usage
Simple model usage with huggingface library:
from transformers import pipeline
model_path = "radlab/polish-roberta-large-v2-qa"
question_answerer = pipeline(
"question-answering",
model=model_path
)
question = "Co b臋dzie w budowanym obiekcie?"
context = """Pozwolenie na budow臋 zosta艂o wydane w marcu. Pierwsze prace przygotowawcze
na terenie przy ul. Wojska Polskiego ju偶 si臋 rozpocz臋艂y.
Dzia艂k臋 ogrodzono, pojawi艂 si臋 r贸wnie偶 monitoring, a tak偶e kontenery
dla pracownik贸w budowy. Na ten moment nie jest znana lista sklep贸w,
kt贸re pojawi膮 si臋 w nowym pasa偶u handlowym."""
print(
question_answerer(
question=question,
context=context.replace("\n", " ")
)
)
with the sample output:
{
'score': 0.3472374677658081,
'start': 259,
'end': 268,
'answer': ' sklep贸w,'
}