Amazon Fake Review Dataset for NLP Model Training

AmazonFakeReview is a dataset for training models to identify fraudulent product reviews. The dataset was created by Mithil and uploaded to Hugging Face in June 2022. It contains text reviews categorized as fake or genuine.

Use Cases

• Train a binary classifier to predict the 'label' column (fake/genuine) from the 'review_text' column.
• Perform sentiment analysis on the 'review_text' to compare linguistic patterns between fake and genuine reviews.
• Extract n-gram or stylistic features from the 'review_text' to build a feature-based detection model.
• Analyze metadata like 'reviewer_id' or 'product_id' for patterns of coordinated fake review campaigns.

Strengths

• Dataset size is categorized as between 10K and 100K rows, providing a substantial sample for model training.
• Includes both tabular structure and raw text modality, suitable for feature engineering and direct NLP applications.

Limitations

• Specific row count, column names, and label balance are unknown, complicating reproducibility and evaluation.
• Data is from 2022 or earlier and may not reflect current review fraud tactics on Amazon.

Provenance

Source

Hugging Face dataset uploaded by author Mithil.

Collection Method

Presumably collected and labeled from Amazon product reviews.

Time Range

null

Freshness

Last updated on Hugging Face in June 2022.

Geography

Platform tags suggest a US (Regionus) focus, but specific geography is unknown.