Initial Training

After you upload data and map attributes in the unified dataset, you or a verifier can begin assigning categories to a sample set of records to train the model to categorize records. When the sample is labeled, you run the Apply feedback and update results job.

To make decisions, a the machine learning model mimics expert behavior by:

• Using all available information in the text to support decisions.
• Leveraging information from multiple attributes.
• Weighting less common words as more important.

Assigning and Verifying Record Categorizations

Both curators and verifiers can assign and verify suggested categorizations. See the following topics first:

• Assigning Records in Categorization Projects
• Filtering Records in Categorization Projects
• Verifying Records in Categorization Projects

You can decide when, and how frequently to run the Update results only or Apply feedback and update results jobs while review and verification is in progress.

Applying Feedback and Updating Categorization Results

The Apply feedback and update results job:

• Trains the model using the latest verified record categorizations.
• Applies the model to all records.
• Calculates record categorization confidences and their average.

To run the Apply feedback and update results job, the following conditions must be met:

• Your project must have one or more unified attributes specified as Included in machine learning.
• You must associate the project with a taxonomy.
• You cannot run any other job at the same time.
• The unified dataset must be up-to-date. See Verifying Record Categorizations.

To run the Apply feedback and update results job:

1. Navigate to the categorized records page by selecting the option to the right of Categories. The name of this page is project-specific and reflects what it is you are categorizing.
2. Choose Apply feedback and update results. See Monitoring Job Status.

Updating Categorization Results

The Update results only job applies the model to all records, and calculates record categorization confidences. You can run the Update results job only after at least one Apply feedback and update results job is complete, or after a model is imported for the project.

To update categorization results:

1. Navigate to the categorized records page.
2. Choose the dropdown arrow on Apply feedback and update results, and then select Update results only. See Monitoring Job Status.