Movie Review Sentences with Sentiment Labels

The Stanford Sentiment Treebank contains 11,855 single sentences extracted from movie reviews, each annotated for sentiment. It includes 215,154 unique phrases from parse trees, enabling compositional sentiment analysis. The dataset was created by StanfordNLP based on earlier work by Pang and Lee.

Use Cases

• Train a text classification model for sentiment classification using the labeled sentences.
• Analyze compositional sentiment effects using the 215,154 unique phrases derived from parse trees.
• Benchmark sequence models on the task of sentiment classification for movie review text.

Strengths

• 11,855 single sentences provide a substantial text corpus for analysis.
• 215,154 unique phrases from parse trees allow for detailed compositional study.
• Dataset is specifically designed for the text classification task of sentiment analysis.

Limitations

• Dataset is monolingual (English) and regionally focused (US), limiting cross-lingual or cultural applications.
• Annotations are crowdsourced, which may introduce label noise or inconsistency.
• The underlying movie review data from 2005 may not reflect contemporary language or sentiment.

Provenance

Source

StanfordNLP, based on the dataset introduced by Pang and Lee (2005).

Collection Method

Sentences extracted from movie reviews, parsed with the Stanford parser, and annotated.

Freshness

Last updated on the platform was 2024-01-04.

Geography

Region is tagged as 'us'.