› Tutorials and Maintenance › Tutorial for performing Extraction using Online Learning › Create a project

Create a project

Create a project (a shared project for simplicity) called "TestOnlineLearning." Add a document type called "Invoices" that includes a few invoices. Pre-train only one vendor's invoice, DataResolutions_01.

Perform the following steps in Transformation Designer.

  1. On the main menu, select Project > New Project.
    New project for Extraction
  2. In the Project Settings dialog box, on the General tab, select Enable Online Learning.
    Project Settings for Extraction
  3. To enable the extraction Online Learning, click Advanced and select Use Extraction Online Learning.
    Advanced Online Learning Options for Extraction
  4. Add a document type, Invoices, with two fields: InvoiceNumber and InvoiceDate. Then add a trainable Invoice Group Locator and assign its subfields to your fields.
    Project tree with a document type for Extraction
  5. Select both fields for TotalAgility to monitor in Validation. If monitored fields are changed, the document will be trained.
    If you do not want all trainable fields to be monitored, you should only select the ones that you want.
    Invoice Number field selected
  6. To set the Invoices document type as the default, so you do not have to set up classification, right-click Invoices and select Default Classification Result.
    Extraction project Tree: Default Classification Result selected
  7. Now pre-train the project with an invoice from the Data Resolutions vendor.
    Pretrain project for extraction
  8. Select the Invoices document type and right-click the invoice (example: DataResolutions_01) to pre-train and add it to the training set for that document type.
    Extract documents

    The document opens in the Edit Document dialog box.

  9. Point and click or lasso the values for Invoice Number and Invoice Date.
    Edit document
  10. Click Add To Training Set and close the dialog box.

    As extraction training is incremental (see Incremental learning versus rebuilding a model), you do not need to select Train Extraction.

    Now you have pre-trained the project with one document from one vendor!

    Note the following about the document:

    • The document resides in the Extraction Training set.

    • The document is being used (Use: Selected). You can select to not use the document and exclude it from the final model.

    • The document will be trained when you rebuild the model (Trained: Yes).

  11. Release the project.
    Release project

    Now you are ready to create a process and generate forms to put the project into production. Continue to Create a process and generate forms.