README.md

metadata

license: cc-by-nc-4.0
datasets:
  - APauli/Persuasive-Pairs
language:
  - en
pipeline_tag: sentence-similarity

Model to score relative persuasive language between pairs

More info about training, evaluation, and use in paper is here: https://arxiv.org/abs/2406.17753

Python:

from transformers import AutoModelForSequenceClassification,AutoTokenizer
import torch
modelname='APauli/Persuasive_language_in_pairs'
model = AutoModelForSequenceClassification.from_pretrained(modelname)
tokenizer = AutoTokenizer.from_pretrained(modelname)

def predict(textA, textB, model,tokenizer):
    encoded_input = tokenizer(textA, textB, padding=True, truncation=True,max_length=256, return_tensors="pt")
    with torch.no_grad():
        logits = model(**encoded_input).logits
    score1=logits.detach().cpu().numpy()
    #flipped
    encoded_input = tokenizer(textB, textA, padding=True, truncation=True,max_length=256, return_tensors="pt")
    with torch.no_grad():
        logits = model(**encoded_input).logits
    score2=logits.detach().cpu().numpy()*(-1)
    score = (score1+score2)/2
    return score

Citation

If you find our dataset helpful, kindly refer to us in your work using the following citation:

@misc{pauli2024measuringbenchmarkinglargelanguage,
      title={Measuring and Benchmarking Large Language Models' Capabilities to Generate Persuasive Language}, 
      author={Amalie Brogaard Pauli and Isabelle Augenstein and Ira Assent},
      year={2024},
      eprint={2406.17753},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2406.17753}, 
}