Generating Suggestions

The machine learning part of the schema mapping process generates mapping suggestions. You can choose how similar attributes for Tamr suggestions should be by using the sliding bar on the Schema Mapping page. Once you perform a small number of mappings manually or import them, select Learn from mappings and then choose Generate mapping suggestions.

Once Tamr generates suggested mappings from machine learning for a particular attribute in the input or unified dataset, you can review the results.

Note: If the similarity score is too high, it may appear that there are no recommendations. In this case, adjust the similarity score on the slider bar.

Adjusting the Similarity Score

Tamr displays only those recommendations that are at least as similar as the number on the similarity slider bar. Adjusting the slider bar behaves as follows:

• If no suggested attributes appear, decrease the percentage on the similarity bar.
• If at least one suggestion exists for an atttribute, an orange lightbulb appears. An attribute name next to the lightbulb indicates this is the only attribute suggestion that Tamr recommends at the similarity score you specified.
• If multiple suggested mappings exist for an attribute at the specified similarity threshold, Tamr indicates that in the number next to the lightbulb. The similarity score for each of the mappings is displayed next to the attribute name in parentheses.

Repeat this process as you accept or reject suggestions, or add new datasets to the project.

Accepting or Rejecting Schema Mapping Suggestions

To accept or reject suggestions:

1. Accept suggestions in one of three ways:

• Hover over the lightbulb next to the unified attribute. Then, hover over each individual attribute in the input dataset that appears under Suggestions and choose Map.
• Hover over the lightbulb next to the attribute in the input dataset. Then, hover over each individual unified attribute that appears under Suggestions and choose Map.
• Select one or more input dataset attributes and choose To Top Suggestion or To Suggestions from the Map drop-down menu.

2. Accept or relabel a number of attributes in the input datasets. Once you accept a suggestion made by Tamr, the lightbulb decreases its counter, or disappears if no suggestions remain. If Tamr is incorrect, override the suggestion and map the attribute to the correct unified attribute.
3. Repeat the process. Choose Learn from Mappings and Generate Mapping Suggestions.