File size: 1,275 Bytes
bdc4e77 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
# YouTube Transcripts Dataset
This dataset contains English transcripts (manual or auto-generated) scraped from top videos of selected YouTube channels. The data includes video metadata and full video transcripts suitable for natural language processing tasks such as language modeling, summarization, and topic analysis.
## Dataset Contents
Each entry in the dataset is a JSON object with the following fields:
- `video_id` (string): The unique YouTube video ID.
- `video_url` (string): Full URL to the YouTube video.
- `title` (string): The video title.
- `views` (int): Number of views the video has.
- `upload_date` (string): The upload date in `YYYY-MM-DD` format.
- `transcript` (string): The full English transcript text of the video.
The dataset is stored in [JSON Lines format](https://jsonlines.org/) (`.jsonl`), where each line is a separate JSON object.
## Notice
**Copyright Disclaimer:**
I do **not** own the copyright to the original video content or transcripts. This dataset is compiled purely for **educational and research purposes** only.
## Usage
You can load this dataset easily using the Hugging Face Datasets library:
```python
from datasets import load_dataset
dataset = load_dataset("ZelonPrograms/Youtube")
print(dataset["train"][0]) |