Dow Jones Newswires Sentiment for Return Prediction

72155736 bytes of supplementary materials accompany a probabilistic framework for extracting sentiment from text. The method uses a joint semi-parametric model on text and ordinal responses, employing word screening and normalized ranks to achieve consistent sentiment ranking without full model estimation. Its application to Dow Jones Newswires demonstrates effectiveness in extracting return-predictive signals.

Use Cases

• Predicting stock returns based on sentiment extracted from Dow Jones Newswires
• Ranking sentiment in new documents using a word screening procedure
• Training a model to handle sparse sentiment signals and complex response distributions
• Reproducing research on text-based return predictability using provided supplementary materials

Strengths

• The dataset includes 72155736 bytes of supplementary materials for reproducing the work
• The framework is explicitly applied and validated on Dow Jones Newswires data

Limitations

• The exact number of rows or documents in the dataset is unknown
• Column names and the specific structure of the underlying data are not provided

Provenance

Source

Dow Jones Newswires

Collection Method

Text mining and probabilistic modeling framework developed by Zheng Tracy Ke

Freshness

Last updated on 2026-06-04