Movie Reviews With Context Drift For Sentiment Analysis

Featuring movie reviews mixed with hotel reviews to simulate context drift, intended for a tutorial. It includes added features such as age, gender, and context, with a made-up timestamp. The dataset size is categorized as between 10,000 and 100,000 rows.

Use Cases

• Train a sentiment classification model on movie reviews to detect sentiment drift when hotel reviews are introduced in the production set.
• Analyze the impact of added context features like age and gender on sentiment prediction accuracy across different review types.
• Simulate and study timestamp-based data drift using the provided made-up timestamp feature in a text classification pipeline.

Strengths

• Dataset size is categorized as 10K<n<100K, providing a substantial number of text records.
• Includes multiple added features such as age, gender, and context for enriched analysis.
• Specifically designed for studying context drift between movie and hotel review domains.

Limitations

• The production set is a mixture of movie and hotel reviews, which may introduce unintended confounding variables for drift analysis.
• Key metadata such as exact row count, column definitions, and file formats are unknown.
• Includes a made-up timestamp, limiting its utility for real-world temporal analysis.

Provenance

Source

Arize AI on Hugging Face, derived from the extended IMDb dataset.

Collection Method

Crafted by mixing a Movie Review Dataset with a Hotel Review Dataset and adding synthetic features.

Freshness

Last updated on 2022-07-01.

Geography

Region is specified as US.