Edit model card

Model Card for Bert2Bert-HunSum-1

The Bert2Bert-HunSum-1 is a Hungarian abstractive summarization model, which was trained on the SZTAKI-HLT/HunSum-1 dataset. The model is based on SZTAKI-HLT/hubert-base-cc.

Intended uses & limitations

  • Model type: Text Summarization
  • Language(s) (NLP): Hungarian
  • Resource(s) for more information:

Parameters

  • Batch Size: 13
  • Learning Rate: 5e-5
  • Weight Decay: 0.01
  • Warmup Steps: 16000
  • Epochs: 15
  • no_repeat_ngram_size: 3
  • num_beams: 5
  • early_stopping: True

Results

Metric Value
ROUGE-1 28.52
ROUGE-2 10.35
ROUGE-L 20.07

Citation

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

@inproceedings {HunSum-1,
    title = {{HunSum-1: an Abstractive Summarization Dataset for Hungarian}},
    booktitle = {XIX. Magyar Számítógépes Nyelvészeti Konferencia (MSZNY 2023)},
    year = {2023},
    publisher = {Szegedi Tudományegyetem, Informatikai Intézet},
    address = {Szeged, Magyarország},
    author = {Barta, Botond and Lakatos, Dorina and Nagy, Attila and Nyist, Mil{\'{a}}n Konor and {\'{A}}cs, Judit},
    pages = {231--243}
}
Downloads last month
20
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.

Dataset used to train SZTAKI-HLT/Bert2Bert-HunSum-1