# uk-parliament-hansard-modern-typical-parquet

Modern UK Hansard debates exported into a **general-purpose (“typical”) Parquet schema** with richer columns (useful outside nanochat/FineWeb-style training pipelines).

## What’s inside

- **Format**: `typical-parquet-v1`
- **Layout**: per-source subfolder(s), each containing `part_*.parquet`, plus a top-level `metadata.json`
- **Columns** (per row): `id`, `source`, `doc_id`, `doc_type`, `title`, `date`, `house`, `speaker`, `text`, `char_count`, `word_count`, `metadata`, `created_at`
- **Parquet settings**: zstd (level 3), `use_dictionary=True`, `write_statistics=True`

## HuggingFace

Dataset repo: `JayJayThrowThrow/uk-parliament-hansard-modern-typical-parquet`

## Loading

```python
from datasets import load_dataset

ds = load_dataset("JayJayThrowThrow/uk-parliament-hansard-modern-typical-parquet", split="train")
print(ds.column_names)
```