11,588 Polish-language product reviews sourced from Allegro.pl categorized by sentiment ratings on a 1-to-5 scale. Each entry consists of a text review of at least 50 words and is organized into pre-defined train, development, and test splits.
Use Cases
- Train sentiment classification models to predict numerical ratings from 1 to 5 based on review text
- Develop natural language processing tools specifically for the Polish language using the provided text corpus
- Evaluate model performance on long-form consumer feedback using the hidden ratings in the test split
Strengths
- 11,588 total reviews extracted from the Allegro.pl e-commerce platform
- Sentiment labels provided as discrete ratings ranging from 1 (negative) to 5 (positive)
- Minimum text length constraint of 50 words per review entry
- Standardized train/dev/test splits for benchmarking sentiment analysis models