Amazon Product Reviews for Multi-Class Text Classification

OpenML hosts a dataset containing 2,000 Amazon product reviews, structured for a 10-class classification task. The dataset includes 100 columns of features, with the sampling process employing stratification to maintain class balance. Its specific creation details, such as author and original collection timeframe, are not provided.

Use Cases

• Train a model to predict the review's assigned class label using the 100 provided feature columns.
• Perform sentiment or topic analysis on the review text column to categorize customer feedback.
• Benchmark stratified sampling techniques on a 10-class text dataset with 2,000 rows.
• Analyze feature importance among the 100 columns for predicting product review categories.

Strengths

• Contains 2,000 individual review records.
• Designed for a 10-class classification problem with 100 feature columns.
• Uses stratified sampling to ensure balanced representation across the 10 target classes.

Limitations

• Sample size of 2,000 rows may be insufficient for training complex deep learning models.
• Unknown column specifics prevent detailed feature engineering insights.
• Lack of metadata like review dates, product categories, or star ratings limits temporal and contextual analysis.

Provenance

Source

OpenML platform.

Collection Method

Sampled from a larger Amazon commerce reviews corpus, with stratification applied to the target variable.