Update README.md

README.md

@@ -1,20 +1,85 @@
----
-license: apache-2.0
-dataset_info:
-  features:
-  - name: file_name
-    dtype: string
-  - name: content_sindhi
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 55664734
-    num_examples: 967
-  download_size: 23097177
-  dataset_size: 55664734
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
----
+---
+license: apache-2.0
+dataset_info:
+  features:
+  - name: file_name
+    dtype: string
+  - name: content_sindhi
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: 55664734
+    num_examples: 967
+  download_size: 23097177
+  dataset_size: 55664734
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+
+
+---
+
+# Sindhi Legal Dataset (Pakistan)
+
+## Dataset Details
+
+### Dataset Description
+
+This dataset contains **Sindhi translations of Pakistani legal documents**, originally written in English.  
+It is intended to support **legal NLP**, **low-resource language research**, and **Sindhi language AI applications**.
+
+The dataset includes constitutional provisions, ordinances, acts, and other statutory legal texts relevant to Pakistan.  
+All documents were translated into **Sindhi (Arabic script)** using a controlled, chunk-based machine translation pipeline to preserve structure and completeness.
+
+- **Curated by:** Danish Mahdi  
+- **Funded by:** Not applicable  
+- **Shared by:** Danish Mahdi  
+- **Language(s) (NLP):** Sindhi (`sd`), source language English (`en`)  
+- **License:** Apache License 2.0  
+
+---
+
+### Dataset Sources
+
+- **Repository:** [Sindhi Legal Dataset on Hugging Face](https://huggingface.co/datasets/DanishMahdi/Sindhi_Legal_2)  
+- **Paper:** Not available  
+- **Demo:** Not available  
+
+---
+
+## Uses
+
+### Direct Use
+
+This dataset is suitable for:
+
+- Sindhi **Legal NLP**
+- **Text generation** and language modeling
+- **Retrieval-Augmented Generation (RAG)** systems
+- Legal **question answering**
+- Fine-tuning **LLMs for Sindhi**
+- Legal document analysis and summarization
+- Low-resource language research
+
+### Out-of-Scope Use
+
+- Legal decision-making or legal advice
+- High-stakes legal interpretation
+- Commercial legal compliance without human verification
+- Tasks requiring human-verified or certified translations
+
+---
+
+## Dataset Structure
+
+The dataset consists of **967 legal documents**, provided as a single `train` split.
+
+Each record has the following structure:
+
+```json
+{
+  "file_name": "example_document.pdf",
+  "content_sindhi": "سنڌي ۾ ترجمو ٿيل قانوني متن..."
+}