Transact

  1. Home
  2. Transact
  3. Features and Functions
  4. Semantik AI Engine
  5. Universal Document Automation

Universal Document Automation

Introduction

Applies to: Transact version 2022.1.00 or newer.

Availability: Windows and Linux on-premise or cloud.

This article describes how to configure and use Universal Document Automation for Ephesoft Transact. Universal Document Automation uses the Semantik Artificial Intelligence (AI) Engine to analyze any document, automatically recognizing and extracting key-value pairs.

This technique differs from the standard Transact approach, where batch class designers build document types to model the documents that a batch class is expected to process. Universal Document Automation uses a batch class, but it only requires a single document type with no required index fields, extraction rules, or validation rules. This single document type can process any document.

Note: See Functional Limitations for more information on what is not currently supported.

This feature supports all languages that are supported by Transact. For a list, see Supported Languages.

Use Cases

This feature allows for document automation without the need to create a template, train the system or configure the solution. It can recognize any document type and identify key-value pair entities for any structured and semi-structured document type out of the box. Universal Document Automation is useful when:

  • You want to extract data from documents without defining the document type.
  • You need to process documents that you haven’t seen before and can’t model using document types (see Configure a Batch Class with the Hybrid Model).

Universal Document Automation is intended to be configured by Transact administrators.

Licensing

Universal Document Automation is included as part of the standard Transact license. There are no additional fees or licensing requirements.

Functional Limitations

  • Universal Document Automation does not extract tables or checkboxes.
  • The output field names and values are determined by the OCR text from the document.
  • Batch classes cannot be encrypted or decrypted.

Prerequisites

Before you configure and use Universal Document Automation:

  • You will need to install Transact version 2022.1.00 or newer.
  • You will need to download and install the Semantik AI Engine. Refer to Semantik AI Engine for steps.
    • Note: The SRE Key-Value Extraction plugin is included in the Semantik AI Engine installation.

Configure Universal Document Automation

Here’s an overview of how to set up Universal Document Automation:

  1. Create a batch class with one new document type.
  2. Configure the Document Assembler plugin to change “Unknown” documents to the Universal document type that you just created.
  3. Add the SRE Key-Value Extraction plugin and configure it as necessary.

Step 1: Configure the Batch Class

  1. From the Batch Class Management screen, select your batch class and click Open.
  2. Create a new document type. You can use any name, but we recommend using “Universal.” For help, see Add New Document Type. Note: Index fields are not required for this document type.

Figure 1. Universal Document document type with no index fields.

  1. Set the Minimum Confidence Threshold value to 0.

Figure 2. Minimum Confidence Threshold set to 0.

Step 2: Configure the Document Assembler Plugin

  1. If not already added, add the DOCUMENT_ASSEMBLER plugin to the Document Assembly module.
  2. Click Deploy.
  3. In the DOCUMENT_ASSEMBLER Plugin Configuration screen, configure the following parameters:
    • Change Unknown Document Type Switch: ON
    • Change Unknown Document To Document Type: [Name of your Universal document type]

Figure 3. Document Assembler plugin configuration.

Step 3: Add and Configure the SRE Key-Value Extraction Plugin

  1. In the Extraction module, add the SRE_KEY_VALUE_EXTRACTION plugin.

Note: For most uses, the plugin order is not important. If you want to use this plugin in combination with other extraction plugins, then it’s best to position it after the other extraction plugins.

  1. Click Deploy.
  2. Configure the SRE_KEY_VALUE_EXTRACTION plugin according to your workflow needs. Refer to the following table for more information on configuration properties.

Figure 4. SRE Key-Value Extraction plugin configuration.

Configurable Property Options Description
Plugin is Enabled
  • YES
  • NO
This property enables the plugin.
List of Document Types to Process (blank for all) N/A This property specifies which document types the plugin will process. If a document type is not specified, the plugin will skip those documents. If left blank, the plugin will process all documents in the batch class.

Note: This option is only needed when you have multiple document types in a batch class and you don’t want to use the Semantik AI Engine for some of them.

Separate the document types by comma with no spaces. For example: DocType1,DocType2,DocType3.

 

If a found field already exists
  • Update the existing value
  • Update if value is blank, else add an alternative value
  • Create a new field with a unique name (a digit will be added to the end of the name)
  • Do not use the duplicate value
The options define what the Semantik AI Engine will do if it finds a field that already exists. This property applies when you’re using the SRE Key-Value Extraction plugin with another extraction plugin on the same document, or if the same key is found more than once on a document. This may result in an overlap of the key-value pairs between the custom fields defined in the document type and what the Semantik AI Engine finds.
Category name of fields for Validation UI N/A This property will group all the fields that the Semantik AI Engine finds on a separate tab in the Validation UI. The input field defines the name of the tab.

If left blank, fields will be added to the default Transact group (Group 1).

  1. Click Apply.

To process your batch instance, continue to Process a Batch Instance with Universal Document Automation.

If you need to add custom document types to your batch class, continue to Configure a Batch Class with the Hybrid Model.

Configure a Batch Class with the Hybrid Model

You can configure a batch class that uses both the Universal document type and custom document types. To configure a batch class using the hybrid model, complete the steps in Configure Universal Document Automation. Then, complete these additional steps:

  1. Create one or more custom document types. For help, see Add New Document Type.

Figure 5. Custom and Universal document types in a batch class.

  1. Configure the new document type(s) following the standard Transact workflow. For help, see Document Type Additional Configurations.

Important: Classification is critical in a batch class that is configured using custom and Universal document types because Transact labels any documents that can’t be classified as “Unknown.” Part of the Universal Document Automation configuration involves automatically changing Unknown documents to the Universal document. If your custom document type is not classified correctly, it will be processed as a Universal document type.

  1. In the SRE_KEY_VALUE_EXTRACTION plugin, configure the following parameter:
    • List of Document Types to Process (blank for all): [Name of your Universal document type]

Note: This setting will send only the Universal document to the Semantik AI Engine, not the other document types.

Process a Batch Instance with Universal Document Automation

  1. From the Batch Class Management screen, click Upload Batch in the left fly-out menu. Alternatively, import your documents using a different method.
  2. Upload your documents.
  3. Select the batch class and click Start Batch.

Note: Universal Document Automation does not need to stop for review or validation because there are no classification confidence levels or validation rules that will stop the workflow. To force the documents to stop for validation, see How to Force Validation.

The extraction results are exported via the standard Transact methods according to your export configuration. For more information on export plugins, see the Export Module documentation. Note: The output values, including dates and numbers, are returned as strings.

How to Force Validation

Universal Document Automation is designed to facilitate straight-through processing without the need to stop for validation. If you want your documents to stop for validation, here is one method to do so:

  1. Add an index field to the Universal document type. You can name this field anything; for this tutorial, we named it “Stop.”
  2. In the Additional Configurations column, select the Force Validation checkbox. Note: Select the Hidden checkbox if you want to hide this field from showing on the Validation screen.

Figure 6. Force Validation and Hidden checkboxes.

  1. Click Apply.

This is an example of using the “Stop” index field to force validation (without the Hidden checkbox selected):

Figure 7. Force Validation example.

Was this article helpful to you? Yes No