Datasetpublic

KhiredNetworks/synthetic-product-reviews

Synthetic Product Reviews A high-quality synthetic dataset of 10,000 product reviews across 5 categories. Dataset Summary Each review was generated by combining realistic review templates with randomly selected products, features, and sentiment-appropriate phrases. The dataset is fully artificial and contains no real user information. Supported Tasks and Leaderboards Sentiment Analysis: Classify sentiment (positive, neutral, negative) Rating… See the full description on the dataset page: https://huggingface.co/datasets/KhiredNetworks/synthetic-product-reviews.

sourceHugging Facemitupdated 2mo agoView on Hugging Face

1likes15downloads

Dataset Card

Synthetic Product Reviews

A high-quality synthetic dataset of 10,000 product reviews across 5 categories.

Dataset Summary

Each review was generated by combining realistic review templates with randomly selected products, features, and sentiment-appropriate phrases. The dataset is fully artificial and contains no real user information.

Supported Tasks and Leaderboards

—Sentiment Analysis: Classify sentiment (positive, neutral, negative)
—Rating Prediction: Predict rating (1-5 stars) from review_text
—Text Generation: Fine-tune models to generate product reviews

Languages

English (en)

Dataset Structure

Data Instances

json

{
  "review_id": 0,
  "category": "Electronics",
  "product_name": "Wireless Earbuds Pro",
  "review_title": "Loving my new Wireless Earbuds Pro!",
  "review_text": "I absolutely love my Wireless Earbuds Pro! The battery life is fantastic. Truly a game changer!",
  "rating": 5,
  "sentiment": "positive"
}

Data Fields

—review_id: unique integer identifier
—category: product category (Electronics, Home & Kitchen, Books, Clothing, Sports)
—product_name: name of the fictional product
—review_title: short review heading
—review_text: full review text (2–5 sentences)
—rating: integer star rating from 1 to 5
—sentiment: derived from rating (positive, neutral, negative)

Data Splits

The dataset is provided as a single train split. You can create your own validation/test splits (e.g., 80/10/10).

Dataset Creation

Curation Rationale

Created to provide a clean, unbiased dataset for sentiment analysis tutorials and small-scale model fine-tuning without privacy concerns.

Source Data

Synthetically generated using Python with fixed random seed (42) for reproducibility.

Additional Information

Licensing Information

MIT License.

Citation Information

If you use this dataset, please cite:

bibtex

@misc{synthetic_product_reviews2024,
  author = {KhiredNetworks},
  title = {Synthetic Product Reviews},
  year = {2024},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/KhiredNetworks/synthetic-product-reviews}}
}