Ananya80/adaption-nyayasetu-legal-assist

This dataset is a remastered version prepared using Adaption's Adaptive Data platform.

adaption-nyayasetulegalassist

This dataset contains multilingual prompt-completion pairs for Indian legal and civic service assistants, covering tasks like document extraction, identity mismatch detection, and legal guidance in languages such as Hindi, Tamil, Marathi, and Punjabi. The completions provide structured JSON responses or empathetic textual advice regarding government certificates, bail applications, and fraud reporting while explicitly disclaiming official authority. Samples demonstrate handling of user facts, required documents, and next steps for various legal scenarios.

Dataset size

There are 54,543 data points in this dataset. This is an instruction tuning dataset.

Quality of Remastered Dataset

The final quality is A, with a relative quality improvement of 4.4%.

Domain

• —Legal (73%)
• —Governance (23%)
• —Technology (4%)

Language

• —English (62%)
• —Malayalam (12%)
• —Hindi (8%)

Tone

• —Helpful (42%)
• —Professional (35%)
• —Cautious (19%)

Evaluation Results

• —Quality Gains: <img src="https://proteus-prod-public.s3.us-east-1.amazonaws.com/temp/5f02a9a9-de74-475c-9ab1-7993acb695cc.png" alt="QualityGains" style="max-width: 50%; display: block; margin-left: auto; margin-right: auto;" />

• —Grade Improvement: <img src="https://proteus-prod-public.s3.us-east-1.amazonaws.com/temp/0665fb3d-4387-44db-a0cc-630d330abe47.png" alt="Grade" style="max-width: 50%; display: block; margin-left: auto; margin-right: auto;" />

• —Percentile Chart: <img src="https://proteus-prod-public.s3.us-east-1.amazonaws.com/temp/a8e49c94-929f-4fdd-93df-05a9d06ab3a2.png" alt="Percentile Chart" style="max-width: 50%; display: block; margin-left: auto; margin-right: auto;" />