README.md

LexDeMod

Release of LexDeMod and Importance annotation Dataset associated with EMNLP 2022 (Agent-Specific Deontic Modality Detection on Legal Language) and EMNLP 2023 (What to Read in a Contract? Party-Specific Summarization of Legal Obligations, Entitlements, and Prohibitions) paper, respectively
Authors: Abhilasha Sancheti, Aparna Garimella, Balaji Vasan Srinivasan, Rachel Rudinger .

1. deontic_data: contains dataset files for the two tasks:

   • Agent-specific Multi-label Classification: The main dataset is in lease_noisy_removed folder and _et_, _est_, _est_masked_ folders contain the files used for ablation experiments in Table 4. _entity, and _entity_random contains files for the generalization experiments in Table 6.

   • Agent-specific Trigger Span Detection: The main dataset is in lease_bio_noisy_removed folder. It contains files used in Table 5. _entity, and_entity_random contains files for the generalization experiments in Table 6.

   • Similar convention is used for the employment and private lease datasets used in the generalization experiments.

   • Each folder contains train.txt, dev.txt, test.txt files corresponding to each split of the dataset. Folders may contain a tenant and landlord folder in addition with text.txt files for the partitioned dataset.

   • test_annotated_data.csv and train_eval_annotated_data.csv contains raw annotations collected from the annotators.

2. importance_data: contains 3 folders

   • ranker: contains train.txt, dev.txt, and test.txt files for the 3 folds used to train and evaluate the importance ranker -- each line contains data in the form [PARTY] sentence1 </s> sentence2 label (0 if sentence1 is more important than sentence2, 1 otherwise)
   • ete: contains ground truth summaries for contracts (end-to-end summarisation) at compression ratios 0.5, 0.10, 0.15 for each fold in the form of a json file. -- each json is a list of jsons for 5 contracts; for each contract a ranked list of obligations (obl), entitlements (ent), and prohibitions (pro) is present both for tenant and landlord
   • categorizer: contains train.txt, dev.txt, and test.txt files for the 3 folds used to train and evaluate the content categorizer -- each line contains data in the form [PARTY] sentence label (list of 1s and 0s indicating the presence or absense of a category)

If you found this repository helpful, please cite our EMNLP 2022, and 2023 papers.

@inproceedings{sancheti-etal-2022-agent,
    title = "Agent-Specific Deontic Modality Detection in Legal Language",
    author = "Sancheti, Abhilasha  and
      Garimella, Aparna  and
      Srinivasan, Balaji Vasan  and
      Rudinger, Rachel",
    booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, United Arab Emirates",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.emnlp-main.795",
    doi = "10.18653/v1/2022.emnlp-main.795",
    pages = "11563--11579"
}

@inproceedings{sancheti-etal-2023-what,
    title = "What to Read in a Contract? Party-Specific Summarization of Legal Obligations, Entitlements, and Prohibitions",
    author = "Sancheti, Abhilasha  and
      Garimella, Aparna  and
      Srinivasan, Balaji Vasan  and
      Rudinger, Rachel",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
}