Amazon Product Reviews Annotated for Complexity and Sentiment

A manually annotated dataset of Amazon product reviews, labeled according to perceived complexity using appraisal theory categories of appreciation and judgment. The dataset was created by Shoukat Ullah and last updated on May 26, 2026. It is used to evaluate a hybrid deep learning model combining DistilRoBERTa and BiGRU.

Use Cases

• Training models to classify product reviews based on user-friendliness and learning curve (appreciation) as described.
• Training models to evaluate products based on effectiveness and appropriateness (judgment) as described.
• Benchmarking NLP models for sentiment analysis against reported state-of-the-art accuracy metrics.
• Applying explainable AI (SHAP) methods to understand model predictions for emotional tendencies and functional effectiveness.

Strengths

• Labels are manually annotated according to a defined theoretical framework (appreciation and judgment).
• The associated model achieved a reported mean fold accuracy of 96.13%.
• The dataset is licensed under CC-BY-4.0, allowing for open use.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (5.5 KB), suggesting a limited scope.

Provenance

Source

Amazon product reviews.

Collection Method

Reviews were manually annotated according to appraisal theory categories.

Freshness

Last updated 2026-05-26 17:43:31; freshness should be verified.