Awesome

Information Extraction for Legal Wills

This repository contains the corpus and code for the paper "Information Extraction from Legal Wills: How Well Does GPT-4 Do?" (To appear in the Findings of the Association for Computational Linguistics: EMNLP 2023).

About The Project

This project introduces a manually annotated dataset for IE from legal wills. Our dataset consists of 16,018 annotations of entities, relations, and events extracted from 45 legal wills from two US states: Tennessee and Idaho. You can find detailed explanation of the information types extracted in our dataset in Appendix B of our paper.

We evaluated GPT-4 with in-context learning on our dataset, in both in-domain (i.e., examples from the same state) and out-of-domain (OOD) (i.e., examples from another state) settings. The evaluation results demonstrate that GPT-4 is capable of handling the legal information extraction task in this in-context-learning setting. However, GPT-4 is not perfect: we observed inconsistent outputs (given a prompt) as well as prompt over-generalization. For more details, please see our paper here.

Usage

Dataset

Our dataset can be used to evaluate large language models on a legal IE task. The raw datasets contain full entities, relations, and events which amount to 16,018 annotations. The datasets used for the LLM evaluation for our project contains four most common entities (i.e., testator, beneficiary, asset, will) and four most common relations (testator-beneficiary, testator-asset, beneficiary-asset, testator-will) only. If you'd like to check the full annotations, please see the raw datasets. If you'd like to evaluate LLM on the legal IE, you may refer to our evaluation datasets.

Codes

A code for OCR (PDF to text)

To extract will texts from PDFs, we ran OCR using python libraries (pdf2image, pytesseract).

To run the code, first install the following libraries by running the code below: pdf2image, pytesseract, tesseract

pip install pdf2image pytesseract tesseract
sudo apt-get install poppler-utils
sudo apt install tesseract-ocr && sudo apt install libtesseract-dev

Once you are finished with installing the libraries, run PDF_to_text.py

A code for transforming raw datasets into evaluation datasets (raw to evaluation datasets)

To transform raw datasets into evaluation datasets, run raw_to_evaluation_datasets.py

Prompt

You can find the prompts used for the LLM evaluation in ./prompt (evaluation_prompt_for_entity.txt, evaluation_prompt_for_relation.txt)

License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. See LICENSE.md for more details.

Paper

If you use this data or code, please cite our paper:

@inproceedings{kwak-et-al-emnlp2023-ie4wills,
    title = "Information Extraction from Legal Wills: How Well Does GPT-4 Do?",
    author = "Alice Kwak and Cheonkam Jeong and Gaetano Forte and Derek Bambauer and Clayton Morrison and Mihai Surdeanu",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
    month = dec,
    year = "2023",
    url = "https://clulab.org/papers/emnlp2023_kwak-et-al.pdf",
    abstract = "This work presents a manually annotated dataset for Information Extraction (IE) from legal wills, and relevant in-context learning experiments on the dataset. The dataset consists of entities, binary relations between the entities (e.g., relations between testator and beneficiary), and n-ary events (e.g., bequest) extracted from 45 legal wills from two US states. This dataset can serve as a foundation for downstream tasks in the legal domain. Another use case of this dataset is evaluating the performance of large language models (LLMs) on this IE task. We evaluated GPT-4 with our dataset to investigate its ability to extract information from legal wills. Our evaluation result demonstrates that the model is capable of handling the task reasonably well. When given instructions and examples as a prompt, GPT-4 shows decent performance for both entity extraction and relation extraction tasks. Nevertheless, the evaluation result also reveals that the model is not perfect. We observed inconsistent outputs (given a prompt) as well as prompt over-generalization."
}

Contact

If you have any questions or comments on our work, please contact the person below.

Alice Kwak - alicekwak@arizona.edu

Acknowledgements

We appreciate Label Studio for providing access to Label Studio Enterprise Cloud Platform through their Academic Program.