Filipino Shopee E-Commerce Reviews with Balanced Sentiment Labels

Filipino-language customer reviews from the Shopee platform, structured for text classification. The dataset contains 10,500 training samples and 2,250 each for validation and testing, with reviews balanced across star ratings from 1 to 5. It is part of the Massive Text Embedding Benchmark (MTEB) and was last updated in May 2025.

Use Cases

• Train sentiment scoring models based on review star ratings.
• Benchmark text classification performance on Filipino-language data.
• Analyze customer feedback patterns in Southeast Asian e-commerce.
• Fine-tune multilingual language models for social media text.

Strengths

• Contains 10,500 training samples, providing a substantial base for model training.
• Data is explicitly balanced, with 2,100 training samples taken for each review star from 1 to 5.
• Includes separate validation and test sets, each with 2,250 samples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The specific source and collection methodology for the reviews are not detailed in the provided metadata.

Provenance

Source

huggingface

Collection Method

Constructed by randomly sampling reviews from the Shopee platform, balanced by star rating.

Time Range

null

Freshness

Last updated 2025-05-06 12:08:46; freshness should be verified.

Geography

Likely focused on the Philippines, given the language and platform.