Yelp Review Full: Up to 1 Million Sentiment-Labeled Business Reviews

Holding between 100,000 and 1,000,000 English business reviews extracted from the 2015 Yelp Dataset Challenge. Created by Yelp, the data pairs review text with sentiment labels to support research in text classification.

Use Cases

• Training sentiment analysis models to predict star ratings from review text
• Fine-tuning language models on consumer-style English text
• Benchmarking text classification algorithms using the Arxiv150901626 methodology

Strengths

• Contains between 100,000 and 1,000,000 records
• Crowdsourced labels from real-world business interactions
• Standardized benchmark cited in Arxiv150901626

Limitations

• Temporal staleness as data is from 2015
• Geographic bias toward the United States
• Monolingual English focus

Provenance

Source

Yelp Dataset Challenge 2015

Collection Method

Crowdsourced

Time Range

Up to 2015

Freshness

Data originates from 2015; last updated on the platform in January 2024.

Geography

United States