Hungarian Sentence-level Sentiment Analysis Model with XLM-RoBERTa

For further models, scripts and details, see our repository or our demo site.

  • Pretrained model used: XLM-RoBERTa base
  • Finetuned on Hungarian Twitter Sentiment (HTS) Corpus
  • Labels: 0 (very negative), 1 (negative), 2 (neutral), 3 (positive), 4 (very positive)

Limitations

  • max_seq_length = 128

Results

Model HTS2 HTS5
huBERT 85.56 68.99
XLM-RoBERTa 85.56 66.50

Citation

If you use this model, please cite the following paper:

@article {laki-yang-sentiment,
      author = {Laki, László János and Yang, Zijian Győző},
      title = {Sentiment Analysis with Neural Models for Hungarian},
      journal = {Acta Polytechnica Hungarica},
      year = {2023},
      publisher = {Obuda University},
      volume = {20},
      number = {5},
      doi = {10.12700/APH.20.5.2023.5.8},
      pages=      {109--128},
      url = {https://acta.uni-obuda.hu/Laki_Yang_134.pdf}
}
Downloads last month
4,067
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.