Open filtering options by selecting Filter on the Pairs page.
To close the filters panel, select Filter again.
Tips for using filters to review and label record pairs effectively:
- Prioritize high-impact pairs.
- For a "To-do" list, set the My assignments filter to Open. Alternatively, set the My response filter to No Response.
- Attribute filters allow you to select attributes and how similar or dissimilar their values are.
- To reduce the number of record pairs that display further, you can use more than one filter.
Filter by high-impact pairs or by pairs with comments.
High-impact record pairs are pairs which have the most utility for training the machine learning model.
Filter by your own assignments or by all assigned pairs.
When you select either filtering option, you can further filter by assignments that are skipped, open, and completed.
Filter by expert, Tamr Core, and your own responses to pair suggestions, as well as by Tamr Core's confidence in its pair suggestions.
Filter by selected source datasets. You can specify the source dataset of each record in a pair.
(see example below of filtering by attribute similarity)
Filter by the similarity of selected attributes and the specified similarity score range.
When Tamr Core pairs two records, it compares the values for each attribute and calculates a similarity score from 0 to 100 for that pair of values. This filter option can help you find and label record pairs that, based on your knowledge of the data, have a greater likelihood of matching or not matching.
Because Tamr Core cannot compute a similarity score if one or both of the values are null, empty, or blank, you can also define a filter for pairs that do not have a similarity score for an attribute. Enable the Pairs without similarity scores toggle.
To remove all filters, select Remove next to Filter .
Updated about 1 month ago