Model description

Cased fine-tuned XLM-RoBERTa-large model for Hungarian, trained on a dataset (~13k sentences) provided by National Tax and Customs Administration - Hungary (NAV): Public Accessibilty Programme.

Intended uses & limitations

The model is designed to classify sentences as either "comprehensible" or "not comprehensible" (according to Plain Language guidelines):

Label_0 - "comprehensible" - The sentence is in Plain Language.
Label_1 - "not comprehensible" - The sentence is not in Plain Language.

Training

Fine-tuned version of the original xlm-roberta-large model, trained on a dataset of Hungarian legal and administrative texts.

Eval results

Class	Precision	Recall	F-Score
Comprehensible / Label_0	0.74	0.65	0.70
Not comprehensible / Label_1	0.71	0.79	0.74
accuracy			0.72
macro avg	0.73	0.72	0.72
weighted avg	0.72	0.72	0.72

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("uvegesistvan/Hun_RoBERTa_large_Plain")
model = AutoModelForSequenceClassification.from_pretrained("uvegesistvan/Hun_RoBERTa_large_Plain")

Citation

Bibtex:

@PhDThesis{ Uveges:2024,
  author = {{"U}veges, Istv{\'a}n},
  title  = {K{\"o}z{\'e}rthet{\"o} és automatiz{\'a}ci{\'o} - k{\'i}s{\'e}rletek a jog, term{\'e}szetesnyelv-feldolgoz{\'a}s {\'e}s informatika hat{\'a}r{\'a}n.},
  year   = {2024},
  school = {Szegedi Tudom{\'a}nyegyetem}
}