Request Access to InLegalLaySum Dataset
We hereby request for the dataset named ‘InLegalLaySum’ which contains Indian Court Case Judgments in English along with their lay summaries. This Dataset is a result of research conducted at Indian Institute of Technology (IIT) Kharagpur and National Institute of Technology (NIT) Hamirpur in India. Please provide the following information and agree to the terms of use before accessing the dataset.
By requesting access, you agree to use this dataset only in accordance with its terms of use.
Log in or Sign Up to review the conditions and access this dataset content.
InLegalLaySum: A Novel Benchmark for Lay Summarization of Indian Legal Documents
InLegalLaySum is a benchmark dataset for lay summarization of Indian legal documents. It is designed to support research on generating lay summaries that are easier for non-expert readers to understand without losing any essential information. The dataset contains 9,570 Indian court judgments from the Supreme Court and High Courts of India, along with expert-written summaries and simplified summaries.
Dataset Overview
Legal judgements are often lengthy and contain specialized legal terminology, complex sentence structures and domain-specific expressions. These characteristics can make legal information difficult to understand for people without legal background. While conventional legal summarization primarily focuses on reducing the length of a legal document, lay summarization additionally focuses on improving the readability and accessibility of the resulting summary for non-expert readers. InLegalLaySum is developed to facilitate research in this direction for Indian legal documents.
The dataset contains:
- 9,570 Indian court judgments.
- Judgements from the Supreme Court and High Courts of India.
- Expert-written English summaries
- Automatically generated simplified summaries for the training set
- Separate training and test splits
Creation of Simplified Summaries
The simplified summaries in the training set are generated using Gemini-2.0-Flash. The generation prompt instructed the model to rewrite the expert-written legal summaries for a high-school reader while preserving the original content and structure. The instructions emphasized:
- Using simple vocabulary.
- Using short and clear sentences where appropriate.
- Preserving the original information.
- Preserving paragraph structure.
- Avoiding unnecessary legal jargon.
- Avoiding the addition of opinions or unsupported information.
- Not introducing additional sections or bullet points.
Dataset Composition
The dataset consists of 7,470 training samples and 2,100 test samples.
| Split | Number of Samples | Contents |
|---|---|---|
| Training | 7,470 | Judgement + Expert Summary + Simplified Summary |
| Test | 2,100 | Judgement + Expert Summary |
| Total | 9,570 |
Dataset Folder Structure
InLegalLaySum/
├── InLegalLaySum_Train
└── <id1>/
├── Judgement.txt # Original case judgement
├── Expert_Summary.txt # Summary of judgement written by law practitioner
└── Simplified_Summary.txt # Summary generated by Gemini-2.0-Flash
├── InLegalLaySum_Test
└── <id1>/
├── Judgement.txt # Original case judgement
└── Expert_Summary.txt # Summary of judgement written by law practitioner
Dataset Download
The InLegalLaySum dataset is available as a gated dataset on Hugging Face.
First, request access from the InLegalLaySum dataset page. After your access is approved, log in to your Hugging Face account:
hf auth login
Then download the complete dataset using:
hf download trishitasaha/InLegalLaySum --repo-type dataset --local-dir InLegalLaySum
Alternatively, the dataset can be downloaded programmatically using Python:
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="trishitasaha/InLegalLaySum",
repo_type="dataset",
local_dir="InLegalLaySum"
)
Terms and Conditions
- Purpose of Use of the Dataset: We agree to utilize the Dataset exclusively for non-commercial academic research purposes and shall refrain from any commercial exploitation, including but not limited to commercial product development, consulting services, or for-profit research endeavours.
- Confidentiality: We shall not share or disclose the Dataset to any third party.
- Publication: If we publish research findings derived from the use of the Dataset, we will include the appropriate citation to the paper “InLegalLaySum: A Novel Benchmark for Lay Summarization of Indian Legal Documents” authored by Trishita Saha, Debtanu Datta, Shubham Tiwari, Mohit Kaushal, Priyank Rathore, Mohammad Khalid Pandit, Pawan Goyal, Saptarshi Ghosh, accepted at Findings of the Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (AACL-IJCNLP Findings), 2026.
- Downloads last month
- 55