---
license: cc-by-4.0
language:
- hu
---

# YouTube Dataset (Semantic Chunks)

Hungarian YouTube comments dataset preprocessed with semantic chunking.

## Stats

| | |
|---|---|
| **Rows** | 629,767 |
| **Tokens** | 21,339,568 |
| **Tokenizer** | `magyar-nlp-szine-java/exotic_modernbert_128k_tokenizer_modified` |

## Columns

- `text` - Chunked text content
- `token_count` - Token count per chunk
- `source_id` - Original source row index
- `chunk_id` - Unique chunk identifier
- `video_id` - Source YouTube video ID
- `channel_title` - YouTube channel title