---
license: mit
task_categories:
  - feature-extraction
  - text-classification
language:
  - en
tags:
  - news
  - embeddings
  - semantic-search
size_categories:
  - 100K<n<1M
---

# News Embeddings Dataset

This dataset contains news article embeddings generated using the [Qwen/Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B) model.

## Source

The original news articles were sourced from the [Global News Dataset on Kaggle](https://www.kaggle.com/datasets/everydaycodings/global-news-dataset?resource=download).

## Dataset Structure

Each row contains:
- `article_id`: Unique identifier for the article from the source dataset
- `text`: Concatenation of title and description (separated by `\n---\n`)
- `embedding`: 1024-dimensional embedding vector

## Embedding Model

- **Model**: Qwen/Qwen3-Embedding-0.6B
- **Prompt**: "Instruct: Given a news article summary, generate a semantic embedding that captures its key features and characteristics \nArticle:"

## Usage

```python
import polars as pl
from huggingface_hub import hf_hub_download

# Download the parquet file
path = hf_hub_download(repo_id="live-wire/news_embeddings", filename="embed.parquet", repo_type="dataset")
df = pl.read_parquet(path)
```