Individual data samples from your dataset or selected data subset are shown, where each sample is a row in the table. In addition to the text, a sample’s associated gold label, predicted label, and DEP score are included as data attribute columns. By default, the samples are sorted by decreasing DEP score.

Fig. The Dataset View

Customization

As shown below, the Dataset View can be customized in the following ways:

  • Sorting by DEP, Confidence or Metadata Columns

  • Filtering to a specific class, DEP range, error type or metadata values

  • Selecting and de-selecting dataset columns

Data Selection

Each row or data sample can be selected to perform an action. As demonstrated in Test Drive Galileo - Movie Reviews, we can easily identify and export data samples with annotation errors for relabeling and/or further inspection. See Actions for more details.