Datasets judgments markdown-1970-2025

Kimi K2
1

Hi everyone,

I’ve compiled a very large private dataset of Pakistan’s legal court rulings in English, covering the years 1970 to 2025. This dataset contains around 190,000 files, all in clean Markdown format.

These documents were obtained from a private, paid legal database, and downloading them required significant time and effort due to the platform’s security measures. Because these rulings are not publicly available, models like ChatGPT, Claude, Gemini, or Grok do not currently have access to this dataset (aside from a few documents that are public).

I would like to request that this dataset be considered for inclusion in future model training, as it could greatly improve legal understanding especially related to Pakistani case law.

Thank you.

download link

thx

2 Likes
2

waiting for reply from admin kindly lookup this data and add in your next model training thanks

3

hey its ijlal im thankful for you contribution. do hit me up on [email protected].

4

Salman i am making a legal ai based on kimi. i want to invite you to come join me in it. dont forget to email me. its [email protected]

5

Dear Msalman,

Thank you so much for reaching out and for sharing this incredible dataset. I’m a technical staff member at Moonshot AI (月之暗面), and I wanted to personally respond to your post.

We have successfully received and downloaded the markdown-1970-2025.zip dataset. Our team has completed an initial review and data processing pipeline, including:

  • Parsing and cleaning all 197,244 court judgment documents across 55 years (1970–2025)

  • Extracting structured metadata (court name, judges, case number, judgment date, outcome, etc.)

  • Converting the entire corpus into a clean, machine-readable format for model training

The dataset is remarkably comprehensive — approximately 2.26 billion characters (~564 million tokens) of English legal text, covering all major Pakistani and Bangladeshi courts. This kind of high-quality, domain-specific data is exactly what helps improve model capabilities in underrepresented legal systems.

We plan to incorporate this data into our upcoming model training cycles. The breadth of coverage — from Supreme Court constitutional rulings to specialized tax and labor tribunals — will meaningfully enhance our models’ understanding of Pakistani and South Asian case law.

We truly appreciate the significant time and effort you invested in compiling this dataset, especially given the access restrictions of the source platform. Contributions like yours are invaluable to advancing AI’s ability to understand and reason about legal texts across diverse jurisdictions.

Thank you again for your generosity and dedication. If you have any additional data or updates in the future, we would be very happy to hear from you.

Warm regards,
Moonshot AI Team

1 Like