concretejungles/20newsgroups-paraphrased

20newsgroups-paraphrased

Paraphrased version of the 20 Newsgroups dataset. The body of each post has been paraphrased while the original email headers (From, Subject, Organization, etc.) are preserved verbatim. Topic labels are unchanged.

Model

Paraphrases were generated using [Qwen/Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507).

Original Dataset

• —Source: scikit-learn 20newsgroups
• —Task: Topic Classification
• —Classes: 20 (0-19 (20 newsgroup categories))

Dataset Structure

Columns

• —text (string): The paraphrased text.
• —label (int): The original label, preserved from the source dataset.
• —original_text (string): The original text before paraphrasing.

How to Use

from datasets import load_dataset

ds = load_dataset("concretejungles/20newsgroups-paraphrased")
print(ds["train"][0])

Generation Details

• —Paraphrase model: Qwen/Qwen3-30B-A3B-Instruct-2507
• —Method: Each example was paraphrased with a dataset-specific prompt designed to preserve the label semantics.
• —Note: Email headers (From, Subject, Organization, etc.) are kept verbatim; only the post body is paraphrased.