by Stanford NLP
IMDB is the Large Movie Review Dataset, a binary sentiment classification dataset hosted by Stanford NLP on Hugging Face. It has 25,000 training and 25,000 test reviews.
DatasetFree
Step 1: Install the datasets library with pip install datasets.
Step 2: Load the data with load_dataset("stanfordnlp/imdb").
Step 3: Use the train and test splits for supervised sentiment classification.
Step 4: Use the unsupervised split if you need unlabeled review text.
Step 5: Cite Maas et al. (2011) when you publish results.
25,000 for training, 25,000 for testing and 50,000 unlabeled reviews in the unsupervised split.
Binary sentiment: 0 for negative and 1 for positive.
A text string and a classification label.
The card lists Maas et al., Learning Word Vectors for Sentiment Analysis, from the ACL-HLT 2011 proceedings.