Filter Records

With this component, users are able to sort the information on a dataset by using different parameters. Filters are different for each task, and they can be used in both Annotation and Exploration modes.

How filters work

To see a description of their components, click here.

In both modes, filters work in a very similar way:

Filtering records in the Annotation Mode

Filtering records can be useful in big datasets, when users need to see and annotate a very specific part of the dataset.

For example, if users are annotating a dataset in a token classification task and they need to see how many records are annotated with one or more labels, they can use the Annotation filter and choose the desired combination.

Filtering records in the Explore Mode

In this case, the use is basically the same, but just for analysis purposes.

Another example would be the following: if in a text classification task with available score, users want to sort records to see the highest or lowest loss, the Sort filter can be used choosing the Score field.

NOTE: There are no filters in the Define label rules, as this mode works with queries.