DALE

DALE: 低リソースの法的 NLP のための生成データ拡張

EMNLP 2023 論文の実装: DALE: Generative Data Augmentation for Low-Resource Legal NLP。

DALE の事前トレーニング済み bart-large モデルはここにあります。事前トレーニングデータはここにあります。

手順:

1. 以下を使用して依存関係をインストールします。

    pip install -r requirements.txt
   

2. 必要なファイルを実行します
   PMI マスキングの場合:

    cd pmi/
   sh pmi.sh <config_name> <dataset_path> <output_path> <n_gram_value> <pmi_cut_off>
   sh pmi.sh unfair_tos ./unfair_tos ./output_path 3 95
   

   BART 事前トレーニングの場合:

    cd bart_pretrain/
   python pretrain.py --ckpt_path ./ckpt_path 
                     --dataset_path ./dataset_path> 
                     --max_input_length 1024 
                     --max_target_length 1024 
                     --batch_size 4
                     --num_train_epochs 10
                     --logging_steps 100
                     --save_steps 1000
                     --output_dir ./output_path
   

   BART 生成の場合:

    cd bart_generation/
   python bart_ctx_augs.py --dataset_name "scotus" 
                     --path ./dataset_path 
                     --dest_path ./dest_path
                     --n_augs 5 
                     --batch_size 4
                     --model_path ./model_path
   
   bart_ctx_augs.py -> BART generation for multi-class data generation.
   bart_ctx_augs_multi.py -> BART generation for multi-label data generation.
   bart_ctx_augs_ch.py -> BART generation for casehold dataset.
   

BibTeX のエントリーと引用情報

私たちの論文/コード/デモが役立つと思われる場合は、私たちの論文を引用してください。

 @inproceedings{ghosh-etal-2023-dale,
    title = "DALE: Generative Data Augmentation for Low-Resource Legal NLP",
    author = "Sreyan Ghosh  and
      Chandra Kiran Evuru  and
      Sonal Kumar  and
      S Ramaneswaran and
      S Sakshi and
      Utkarsh Tyagi and
      Dinesh Manocha",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Sentosa, Singapore",
    abstract = "We present DALE, a novel and effective generative Data Augmentation framework for lowresource LEgal NLP. DALE addresses the challenges existing frameworks pose in generating effective data augmentations of legal documents - legal language, with its specialized vocabulary and complex semantics, morphology, and syntax, does not benefit from data augmentations that merely rephrase the source sentence. To address this, DALE, built on an EncoderDecoder Language Model, is pre-trained on a novel unsupervised text denoising objective based on selective masking - our masking strategy exploits the domain-specific language characteristics of templatized legal documents to mask collocated spans of text. Denoising these spans help DALE acquire knowledge about legal concepts, principles, and language usage. Consequently, it develops the ability to generate coherent and diverse augmentations with novel contexts. Finally, DALE performs conditional generation to generate synthetic augmentations for low-resource Legal NLP tasks. We demonstrate the effectiveness of DALE on 13 datasets spanning 6 tasks and 4 low-resource settings. DALE outperforms all our baselines, including LLMs, qualitatively and quantitatively, with improvements of 1%-50%."
}