Published 2011
| Version v4
IMDb Movie Reviews Dataset
Creators
- 1. Stanford University
Description
Description
The IMDB Movie Reviews dataset comprises 50 000 sentiment‑labeled reviews scraped from IMDb (25 000 for training and 25 000 for testing) plus an additional 50 000 unlabeled reviews for unsupervised or semi‑supervised experiments. Reviews are included only when the IMDb rating is strongly positive (≥ 7/10) or strongly negative (≤ 4/10), and each entry is provided in both raw‑text and pre‑processed bag‑of‑words formats, making it a challenging yet unambiguous benchmark for binary sentiment classification. Since its release the dataset has become a standard NLP benchmark, integrated into Hugging Face Datasets, TensorFlow Datasets, Keras, and PyTorch‑NLP.
Please cite the following if you use the data:
Maas, A. L., Daly, R. E., Pham, P. T., Huang, D., Ng, A. Y., & Potts, C. (2011). Learning Word Vectors for Sentiment Analysis. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (pp. 142‑150). Association for Computational Linguistics. https://www.aclweb.org/anthology/P11-1015
Variables
| Name | Description |
|---|---|
| id | Review ID |
| rating | Star rating of a review on a 1-10 scale |
| pos / neg / unsup | The polarity of the review (positive / negative). No polarity score is provided for reviews for unsupervised learning. |
| text | Review text |
Details
| Resource type | Open dataset |
| Title | IMDb Movie Reviews Dataset |
| Creators |
|
| Publisher | Stanford University |
| Year of publication | 2011 |
| Research fields | Business Administration Economics Psychology Sociology Political Science Economic & Social History Communication Sciences Educational Research Other |
| Size | 0.08 GB |
| External resource | https://ai.stanford.edu/~amaas/data/sentiment/ |