Skip to content

Dataset · stanfordnlp

SST-2 Sentiment Corpus

stanfordnlp/sst2

11,855 movie review sentences with sentiment labels for classification and composition analysis.

What's inside

SST-2 is a fully labeled parse tree corpus for sentiment analysis, based on Pang and Lee (2005): 11,855 individual movie review sentences parsed by the Stanford parser. The trees contain 215,154 unique phrases.

What you can do with it

  • Train a sentiment classifier
  • Fine-tune BERT or RoBERTa for SST-2
  • Benchmark and compare models
Tasks
Text classification
Languages
English
Size
10K–100K rows
Downloads this month
45.8K
Likes
179

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("stanfordnlp/sst2")