SlavicNLP/lrec2026-persuasion-sentence-classifier-polish

Model Description

This model performs fine-grained persuasion technique classification for Polish. It is a custom Small Language Model (SLM) trained to identify and categorize specific rhetorical and persuasion strategies used in text.

The model is associated with the paper:

A Corpus of Persuasion Techniques in Slavic Languages
Submitted to LREC-COLING 2026

Model usage

from transformers import pipeline

model = "SlavicNLP/lrec2026-persuasion-sentence-classifier-polish"
classifier = pipeline(task="text-classification", model=model, device=device, top_k=None)

texts = ["To uwłaszczona miłość matki i ojca, u których będzie rosnąć nie nasze, własne dziecko, ale dusza przekazana na przechowanie."]
outputs = classifier(texts, padding=False, truncation=True, max_length=512)
print([output for output in outputs[0] if output["score"] > 0.5])

Expected output (formatted for readability):

[{'label': 'Loaded_Language', 'score': 0.9126505255699158},
 {'label': 'Appeal_to_Values', 'score': 0.9099323153495789}]

Maintainer

Michał Marcińczuk (marcinczuk@gmail.com)