Mbart Summarization Fanpage

🚀 mbart-summarization-fanpage

このモデルは、要約生成のためにFanpageデータセットでfacebook/mbart-large-cc25をファインチューニングしたバージョンです。

以下の結果を達成しています：

• 損失: 2.1833
• Rouge1: 36.5027
• Rouge2: 17.4428
• Rougel: 26.1734
• Rougelsum: 30.2636
• 生成長: 75.2413

🚀 クイックスタート

モデルの使用方法

from transformers import MBartTokenizer, MBartForConditionalGeneration
tokenizer = MBartTokenizer.from_pretrained("ARTeLab/mbart-summarization-fanpage")
model = MBartForConditionalGeneration.from_pretrained("ARTeLab/mbart-summarization-fanpage")

トレーニングのハイパーパラメータ

トレーニング中に使用されたハイパーパラメータは以下の通りです：

• 学習率: 5e-05
• トレーニングバッチサイズ: 1
• 評価バッチサイズ: 1
• シード: 42
• オプティマイザ: Adam (betas=(0.9,0.999), epsilon=1e-08)
• 学習率スケジューラのタイプ: linear
• エポック数: 4.0

フレームワークのバージョン

• Transformers 4.15.0.dev0
• Pytorch 1.10.0+cu102
• Datasets 1.15.1
• Tokenizers 0.10.3

📚 ドキュメント

引用

詳細と結果については公開された研究を参照してください。

@Article{info13050228,
    AUTHOR = {Landro, Nicola and Gallo, Ignazio and La Grassa, Riccardo and Federici, Edoardo},
    TITLE = {Two New Datasets for Italian-Language Abstractive Text Summarization},
    JOURNAL = {Information},
    VOLUME = {13},
    YEAR = {2022},
    NUMBER = {5},
    ARTICLE-NUMBER = {228},
    URL = {https://www.mdpi.com/2078-2489/13/5/228},
    ISSN = {2078-2489},
    ABSTRACT = {Text summarization aims to produce a short summary containing relevant parts from a given text. Due to the lack of data for abstractive summarization on low-resource languages such as Italian, we propose two new original datasets collected from two Italian news websites with multi-sentence summaries and corresponding articles, and from a dataset obtained by machine translation of a Spanish summarization dataset. These two datasets are currently the only two available in Italian for this task. To evaluate the quality of these two datasets, we used them to train a T5-base model and an mBART model, obtaining good results with both. To better evaluate the results obtained, we also compared the same models trained on automatically translated datasets, and the resulting summaries in the same training language, with the automatically translated summaries, which demonstrated the superiority of the models obtained from the proposed datasets.},
    DOI = {10.3390/info13050228}
}