Amazon Commerce Reviews for Multi-Class Text Classification

2000 rows of Amazon product reviews, stratified across 10 distinct classes, intended for text classification tasks. The dataset is sourced from OpenML and is associated with platform tags for sentiment analysis and product categorization. Specific temporal coverage, author, and collection method are not provided.

Use Cases

• Predict the review's product category (nclasses_10) from the review text content.
• Perform sentiment analysis on review text to classify sentiment polarity.
• Train a model for text classification using the stratified 10-class labels to ensure balanced representation.
• Analyze textual features within the 100 provided columns (ncols_100) for natural language understanding tasks.

Strengths

• Contains 2000 review instances (nrows_2000) for model training.
• Stratified sampling (stratify_True) across 10 classes (nclasses_10) helps mitigate class imbalance.
• Includes 100 columns (ncols_100), suggesting a potentially feature-rich representation of the text data.

Limitations

• Sample size of 2000 rows may be insufficient for training complex deep learning models without augmentation.
• Unknown column structure and specific features limit immediate understanding of data utility.
• Lacks metadata on review dates, product domains, and reviewer demographics, restricting temporal and demographic analysis.

Provenance

Source

OpenML platform, originally from Amazon commerce reviews.

Collection Method

Sampled and stratified subset (seed_2) of a larger Amazon reviews corpus.