Upload README.md with huggingface_hub

README.md

@@ -4,37 +4,105 @@ dataset_info:
   - name: example_html
     dtype: string
   - name: expected_json
-    struct:
-    - name: author
-      dtype: string
-    - name: favicon
-      dtype: string
-    - name: id
-      dtype: string
-    - name: image
-      dtype: string
-    - name: published_date
-      dtype: string
-    - name: text
-      dtype: string
-    - name: title
-      dtype: string
-    - name: url
-      dtype: string
+    dtype:
+      url: string
+      title: string
+      text: string
+      author: string
+      published_date: string
+      image: string
+      favicon: string
+      id: string
   splits:
   - name: train
-    num_bytes: 32186875
     num_examples: 450
   - name: test
-    num_bytes: 2516130
     num_examples: 50
-  download_size: 7746196
-  dataset_size: 34703005
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-  - split: test
-    path: data/test-*
+task_categories:
+- text-generation
+- question-answering
+language:
+- en
+size_categories:
+- n<1K
 ---
+
+# CrawlerLM: HTML-to-JSON Dataset
+
+A synthetic dataset for training language models to extract structured JSON data from raw HTML.
+
+## Dataset Description
+
+This dataset contains HTML pages paired with their structured JSON representations, designed for fine-tuning small language models for web scraping and information extraction tasks.
+
+### Dataset Structure
+
+Each example contains:
+- `example_html`: Raw HTML content from real web pages
+- `expected_json`: Structured extraction with fields:
+  - `url`: Page URL
+  - `title`: Page title
+  - `text`: Main text content
+  - `author`: Author name (or null)
+  - `published_date`: Publication date (or null)
+  - `image`: Main image URL
+  - `favicon`: Favicon URL
+  - `id`: Unique identifier
+
+### Data Splits
+
+- **Train**: 450 synthetic variations
+- **Test**: 50 synthetic variations
+
+### Data Sources
+
+- Base HTML samples from Common Crawl
+- Structured extractions via Exa API
+- Synthetic variations generated programmatically
+
+### Use Cases
+
+- Fine-tuning small models for web scraping
+- Training HTML-to-JSON extraction models
+- Benchmarking structured data extraction
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("espsluar/crawlerlm-html-to-json")
+
+# Access splits
+train_data = dataset["train"]
+test_data = dataset["test"]
+
+# Example
+example = train_data[0]
+print(f"HTML length: {len(example['example_html'])} chars")
+print(f"Title: {example['expected_json']['title']}")
+```
+
+## Dataset Creation
+
+Generated using the CrawlerLM pipeline:
+1. Sample diverse URLs from Common Crawl
+2. Filter for quality (SPA detection, content scoring)
+3. Extract structured data via Exa API
+4. Generate synthetic variations (wrappers, noise, perturbations)
+
+## License
+
+MIT
+
+## Citation
+
+```bibtex
+@misc{crawlerlm2025,
+  author = {Jack Luar},
+  title = {CrawlerLM: HTML-to-JSON Dataset},
+  year = {2025},
+  publisher = {HuggingFace},
+  howpublished = {\url{https://huggingface.co/datasets/espsluar/crawlerlm-html-to-json}}
+}
+```