Update README.md

README.md

@@ -21,4 +21,53 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+size_categories:
+- 1K<n<10K
+task_categories:
+- image-to-text
+- image-text-to-text
+- image-text-to-image
+language:
+- ar
+tags:
+- arabicbooks
+- shamelabooks
 ---
+
+# Taqreeb Fatawa wa Rasa'il Ibn Taymiyyah (OCR Dataset)
+
+This dataset is a structurally aligned Arabic Optical Character Recognition (OCR) dataset. It represents the first batch of the `ocr_arabic_books` initiative, featuring the book **Taqreeb Fatawa wa Rasa'il Shaykh al-Islam Ibn Taymiyyah** (تقريب فتاوى ورسائل شيخ الإسلام ابن تيمية) compiled by Ahmad bin Nasir al-Tayyar in 5 volumes.
+
+All 3,537 pages have been systematically processed, normalized, and mapped sequentially to preserve the physical layout of the printed text.
+
+## Dataset Structure
+
+The dataset contains the following fields for each page record:
+
+| Field Name | Type | Description |
+| :--- | :--- | :--- |
+| `image` | `Image` | Scanned page image (JPEG format, 150 DPI) |
+| `page_header` | `string` | Spatially aligned running headers containing Eastern Arabic page numbers and active section titles. |
+| `body` | `string` | Clean Arabic main body text (fully stripped of HTML tags, normalized to Unix newlines, and free of database anchors). |
+| `footnotes` | `string` | The corresponding footnotes belonging to the bottom of the page (`null` if none exist). |
+
+---
+
+## How to Use
+
+The dataset is stored in native Hugging Face Parquet format with embedded image bytes, allowing you to load or stream it with a single line of Python code:
+
+```python
+from datasets import load_dataset
+
+# Load the entire dataset
+dataset = load_dataset("freococo/ocr_arabic_books", split="train")
+
+# Access the first page record
+first_page = dataset[0]
+print(first_page["page_header"])
+print(first_page["body"])
+
+# The image is automatically decoded as a PIL Image object
+first_page["image"].show()
+```