Training by project administrators
A project administrator can train the program using some training batches during project setup. When operators make corrections to recognized documents, the corrected documents will be put into the training batches used by the administrator. The administrator can also make additional changes to the training batches and retrain the program if training by operators does not deliver the desired results.
Once the training is complete, the administrator can lock specific batches to prevent them from being modified by processing operators. To lock a training batch used for a specific company, right-click the batch and then click Lock Training by Operators on the shortcut menu.
Training is only available if ABBYY FlexiCapture can reliably identify the issues by finding the corresponding record in a database.
If you have no databases but still want to use field training, you can accumulate company information by adding records to your data sets while capturing documents. For more information, see Looking up vendors and business units in the database.
Note: The actions described in this section can only be performed by the administrator or by users who have the permissions to set up projects in ABBYY FlexiCapture for Invoices.
The training batch view displays all the document images that will be used for training. To switch to the training batch view, click (Open Field Extraction Training Batches) on the toolbar. Alternatively, click Field Training → Open Field Extraction Training Batches.
The program will display a list of training batches for all the document variants that have been added by the operators. For convenience, important company parameters are displayed in this view, including VATIDs, IBANs, and database IDs.
You can work with document variants in this view.
Managing sample images used for training FlexiLayout variants
Double-click a training batch or right-click a training batch and then click Open on the shortcut menu. A list of sample images used for the corresponding FlexiLayout variant will be displayed.
Each sample image may be in one of the following three states:
- Unused - This sample image is not used for training or testing this FlexiLayout variant.
- For training - This sample image is used for training and testing this FlexiLayout variant.
- For testing - This sample image is used only for testing this FlexiLayout variant.
In this view, you need only one sample image to start training a FlexiLayout variant. However, we recommend having at least three sample images in For training state.
If there is not enough sample images for training, load more images by clicking Load Images.... By default, any newly added images will be marked as For training.
If the images you add are not recognized automatically, force-start the recognition process by clicking Recognize.
- Change the state of the images that you wish to use to For training. To change the state, select the images you need, right-click them, and then select Set Document State → For training. We recommend changing the state of all the sample images to For training if you have not trained this document variant before.
- Check the markup of the all regions and make any adjustments as may be necessary. This is important, because the positions of all the regions will be taken into account.
- To start training the FlexiLayout variant, either
- Select Fields Training → Train from the main menu, or
- Right-click anywhere in the list and select Train from the shortcut menu, or
- Switch into the training batch view and select Train from the shortcut menu of the document variant.
Training progress will be indicated in the Train Document Definition window. As a result of this training, a FlexiLayout variant will be created. Next, the FlexiLayout Variant will be tested on all documents marked For training and For testing. The program will apply the FlexiLayout to each sample image (the user will not see this process) and compare the result with the reference layout (i.e. the layout created manually by the user). Then the quality of matching will be established.
If the new FlexiLayout variant is better than the previous version of this FlexiLayout variant and the main FlexiLayout, the new FlexiLayout variant will be saved and used for subsequent data capture.
If you think that some of the sample images adversely affect the training but you still want to use them for testing, change their state to For testing.
Make sure than at the time of training the regions of all fields on these sample images are located correctly. The testing process itself will not be visible to the user.
Exporting FlexiLayout variants to FlexiLayout Studio for manual adjustment
Right-click a document variant and select Export Trained FlexiLayout... from the shortcut menu.
You may need to do this if you can't achieve the desired results by just adding or removing sample images.
Importing FlexiLayout variants from FlexiLayout Studio
Right-click a document variant and select Import FlexiLayout... from the shortcut menu.
You may need to do this if you have a FlexiLayout that has been created manually in FlexiLayout Studio for the document variant.
Prohibiting further training on documents from a specific company
Right-click a document variant and select Lock Training by Operators from the shortcut menu.
You may want to prohibit training if the FlexiLayout for the given document variant has already been trained (or created manually) and no further training will be necessary during processing.
Prohibiting the use of training results when processing documents from a specific company
Right-click a document variant and select Disabled from the shortcut menu.
Select this option if you want only the main FlexiLayout to be used when processing documents from the given company.
Creating a new training batch
You may want to create a new training batch if you want to train a FlexiLayout variant before operators start capturing documents.
- In the training batch view, right-click an empty field and select New Batch... from the shortcut menu.
After the training batch has been created, choose its associated Document Definition (if your project uses multiple Document Definitions) and the company for which the training results will be used.
- If there is more than one Document Definition in your project, select the desired Document Definition in the Select Document Definition dialog box (right-click the batch and click Change Document Definition... on the shortcut menu).
- Choose the issuing company for the training batch.
Note: Be sure to recognize documents in a batch using only the Document Definition assigned to that batch.
Use the shortcut menu of the batch to change the company or the Document Definition associated with the batch.
The Clear Document Definition command removes the Document Definition associated with the batch. Batches that have no Document Definitions associated with them cannot be used for training.
- Add documents to the training batch by right-clicking it and selecting the Load Images... from the shortcut menu. The newly added documents will be automatically recognized.
- Check whether field regions have been detected correctly and adjust them if necessary.
- Select the documents you want to use for training the FlexiLayout variant, right-click one of them and click Set Document State → For training.
- Start training the FlexiLayout variant. There are several ways to do this:
- In the main window, open the Fields Training menu and click Train;
- Right-click an empty spot in the list and click Train on the shortcut menu.
- Open the training batch view, right-click a FlexiLayout variant and click Train on the shortcut menu.
Note: Disable the Use own recognition settings of FlexiLayout option. This will synch the full-text recognition settings and speed up document processing.
If you wish to import into your training batch a modified or a completely new FlexiLayout to be used for one specific company, complete the following steps:
- In ABBYY FlexiLayout Studio, click File → Export... and in the Export FlexiLayout dialog box that opens, specify the name of the file and the folder where it should be placed.
- In ABBYY FlexiCapture, click Train Field Extraction → Import FlexiLayout...
If you are using the clustering feature, please note the following limitations:
- If you are creating a new FlexiLayout manually, make sure that it covers all the possible document variants originating from the given company—you cannot manually create a FlexiLayout for one cluster only.
- Only a FlexiLayout for the main document fields will be exported. No FlexiLayout can be generated and exported for line item fields, as this type of field uses a separate machine learning algorithm, whose results cannot be exported or modified. However, you can still create a FlexiLayout for line item fields manually.
- Only the FlexiLayout trained for the first cluster will be exported.
- After you import a new or modified FlexiLayout into your training batch:
- There will be no training while processing documents.
- Clustering will be disabled for this batch.
- The imported FlexiLayout will be used for processing all documents from this company, regardless of their cluster.
For details, see Training with clustering.