Train a classifier

You can enhance your automations by training your own classification models, a process that is both versatile and efficient. For detailed instructions on how to proceed, follow the steps from this page.

1. Open your Document Understanding^TM modern project.
2. From the Build section, upload your documents and classify them:
   • To classify the documents automatically, follow the steps described in the Out of the box document types procedure, on the Classify document automatically page.
   • To upload your document to a custom document type, follow the steps described in the Custom document types procedure, on the Classify document automatically page.
   备注：

   • The classifier will be trained using data from uploaded documents.
   • You need at least five documents for each document type to train a classifier.
   • You need at least two document types to train a classifier.

   
   

   

   
   
3. Once you've uploaded your documents, navigate to the Measure section to assess the performance of your classifier.
   提示：

   • Check the recommendations on the right side of the window to improve your classifier's performance.
   • Only document types with at least five documents are displayed in the Metrics tab.
   • Each time you move a document to a different document type, the model is retrained.

   
   

   

   
   
4. Once you are satisfied with your classifier's metrics, proceed to the Publish section.
5. Create a new project version to include the classifier.
   1. Add a name for your project version.
   2. [Optional] Add a description for your project version.
   3. Select only Classifier from the Models to include drop-down list.
6. Deploy the project version, which includes the classifier, to use it in your automations.
   Tip: For more information on how to deploy a project version and automate your process, check the Project version section on the Publish page.