Home BlogDataset Training Counterfactual Statement Detection Models

Training Counterfactual Statement Detection Models

by Morgan Stevens
by
Binary code

Amazon has created a dataset of sentences to train language processing models to identify counterfactual statements. Counterfactual statements, which are statements that take the form “if p were true, then q would be true,” can mislead product retrieval systems. The dataset consists of sentences from customer reviews in English, German, and Japanese, annotations, and words commonly found in counterfactual statements, such as “wished” and “except.” 

Get the data.

Image credit: Flickr user CyberHades

You may also like

Show Buttons
Hide Buttons