Re16: Read the paper ILDC for CJPE: Indian Legal Documents Corpus for Court Judgment Prediction and Explanation

Gods are silent-personal CSDN blog directory

Paper name: ILDC for CJPE: Indian Legal Documents Corpus for Court Judgment Prediction and Explanation
Paper ACL official download address: https://aclanthology.org/2021.acl-long.313/ (The website contains the official ACL video. I'm really deaf with an Indian accent, I can't understand it at all, I'm speechless. This article is not fromThe pictures of the paper are all taken from this video)
ArXiv download address of the paper: https://arxiv.org/abs/2105.13562
Official GitHub project: Exploration-Lab/CJPE

This article is the 2021 ACL paper.The task CJPE proposed in this paper has two requirements: First, predicting the court decision based on the factual description text of the case, that is, whether to support the plaintiff's claim (allowed or dismissed), is a binary classification problem.Second, provide interpretable text for this result.This paper proposes a new English dataset ILDC, from India, English for this task.The input is all case documents except the judgment result, and the output is whether the court supports the plaintiff's claim, a binary classification task.Explanatory texts (manually annotated) of the decision results are provided on the test set.

This blog post is only a brief overview, specific details to be added.

1. CJP

1.1 Model

1.2 Experiment

(The results of each experiment are written directly for the models with very different experimental results) (There is also a table of detailed hyperparameters for each model)

LR: Logistic Regression

2. Interpretation generation

Using XLNet+BiGRU as above:

p>