IQ Bot and Document Automation v.26 release

Review what's new and changed, and the fixes and limitations in IQ Bot and Document Automation for the v.26 release.

Document Automation

What's new
Improve extraction by providing changes to the Validator

You can now improve extraction by providing learning instances with validation feedback for Automation Anywhere models. You can achieve this by enabling the Validation feedback option in the Create learning instance and Edit learning instance windows.

Improve extraction accuracy through validation

Search for form or table fields

When creating a learning instance, when you are configuring fields, you can now search for fields by field name, field label, or data type.

Process invoice-like documents with user-defined document type
You can now process documents that are visually similar to invoices, such as purchase orders and sales orders, which contain key-value pairs and a table structure. You can achieve this by create learning instances with the user-defined document type, using which you can also achieve the following results:
  • Extract from English, Dutch, French, German, Italian, Portuguese (Brazilian), and Spanish language documents.
  • Define exactly which values to extract from a field, such as when a date of birth field also contains the age.
  • Use the Address data type for form fields, which identifies and extracts the structure of an address.
Enhancements to the Validator
  • The Validator now shows the group number for documents processed in a learning instance connected from Automation 360 IQ Bot
  • You can now configure a list of valid values when creating learning instance.
  • When you click a field that requires validation, a message now appears with the confidence level determined by the system and the threshold set in the learning instance.
Monitor document processing metrics in Bot Insight dashboards

You can now visualize data, such as the number of documents uploaded per day, extraction time per document by learning instance, and STP percentage by learning instance, by using the customizable dashboards in the Document Workspace tab.

Document Workspace dashboard

Document Automation now supports the bring your own key (BYOK) use case for the Google Document AI product license

Users can now configure their Control Room environment with Google Document AI licenses purchased from Google. This allows the learning instance to send documents to Google Document AI for extraction and for the Control Room to track consumption in the All licenses page.

Configure key for Google Document AI | Automation 360 licenses

View Document Automation user activities in the audit log

With the AAE_Admin role or a custom role with the View audit log permission, you can now view the status of events related to learning instances, domains, and documents.

Audit log

Create and train standard forms within the Control Room

Users with the AAE_IQ Bot Admin or AAE_IQ Bot Services roles can now create new standard forms in Document Automation. It is no longer necessary to leave the Control Room and open a third-party service to create or modify a standard form. The interface to configure the extraction model is launched from the Learning instances page.

Note: To process documents in a learning instance which uses the standard forms model in Document Automation:
  1. Create the standard form in Document Automation.
  2. Create and train the learning instance in Automation 360 IQ Bot.
  3. Connect the learning instance to Document Automation.

Create a custom extraction model using Standard Forms

What's changed
To connect a learning instance from Automation 360 IQ Bot, you must have one of the following roles:
  • AAE_IQ Bot Admin
  • AAE_IQ Bot Services
  • A custom role with the Connect learning instance permission
Uploaded documents and extracted data are now removed after 90 days even if a document has not completed processing, except in the following cases:
  • If data is downloaded and deleted, the source documents are immediately deleted.
  • If the document is manually validated, the document is stored for 90 days from the date of extraction.
If a document is renamed to a new file extension and uploaded to a Google Document AI pretrained model, the document will fail processing.

Workaround: Rename the document to the correct file extension.

Fixes
Beside the Validate documents link, the value that reflects the number of documents awaiting validation is no longer hidden, and you can now view it without adjusting the width of the Actions column.

Previously, you had to adjust the width of the Actions column to see the value.

You can now successfully draw a box, zoom in, fit the box to match the screen size, and then redraw the box without any issue.

Previously, when you drew a box, zoomed in, clicked Fit to screen, and then tried to redraw the box, the box appeared at a different location from where you clicked.

In the Validator, you can now adjust the system-identified region (SIR) by dragging the corners of the SIR when you zoom in or zoom out.

Previously, you could drag the corners of the SIR but not when you zoomed in or zoomed out.

When you connect to a learning instance from Automation 360 IQ Bot, if the learning instance has an optional field with a set default value and some fields that are unmapped, when you process documents in that learning instance in Document Automation, the default value now appears only in the mapped fields.

Previously, the default value appeared incorrectly in the unmapped fields.

In Community Edition, when you try to delete a learning instance, the following, correct message is now displayed and you can successfully delete the learning instance: Community edition users are limited to 5 learning instances

Previously, although you could delete the learning instance, the following, incorrect message was displayed: Community edition users are limited to 5 learning instances and will not be able to delete.

Fixed an issue where Document Automation sometimes failed to extract data from tables that span across multiple pages. Document Automation now supports data extraction from tables that span across multiple pages for Automation Anywhere models.
You can now use the Document Extraction > Download action in a standalone bot that is not part of an Automation Co-Pilot process.

Previously, this action could be used only in the Downloadbot within the Automation Co-Pilot process that was automatically generated when a user created a learning instance.

After the Validator times out because of inactivity, when you refresh it, instead of showing the most recent document that you validated, the Validator shows the first document that you validated.
Limitations
Formula validation does not support numbers that contain delimiters. For example, 1,000 and 2.0 are not supported.
In some cases, the Bot Agent does not clear out its resources folder. Until the disk space is freed, the production documents in processing will be affected. This issue occurs intermittently.

Workaround: Based on the issue, you must create a task scheduler to clean out the resources folder every day or every week.

There is no notification when the product license is about to expire. When the product license expires, the Learning Instances tab is removed from the Control Room.
Learning instances that use Google Document AI models do not reliably extract all the fields in a table, if the table that spans multiple pages.
In On-Premises installation in Linux environments, document processing sometimes fails because of a storage service error.
Workaround: In the command prompt, run the following script:
cd /opt/automationanywhere/automation360/appdata/Server Files/
sudo chown -R crstorage:controlroom storageservice
Document Automation does not support Oracle Database for On-Premises installations.
Document Automation supports only the following standard ports for On-Premises installations:
  • For HTTP: Port 80
  • For HTTPS: Port 443
Sometimes, files uploaded to train a standard form appear corrupted although the data is still usable.
Standard forms in Document Automation is not supported in a multi-tenant Cloud Control Room that was upgraded from v.25. This limitation does not occur on a fresh installation of a multi-tenant Cloud Control Room.
Learning instances with the user-defined document type do not support Dutch language documents.
The first time you click a learning instance connected from Automation 360 IQ Bot, you are redirected to the IQ Bot homepage instead of the details page for that learning instance.
Formula validation is not supported for custom fields if the field name contains a space.
When you import a learning instance created in a previous version of Document Automation, the check box to send validation feedback to the system is unselected by default. You must edit the learning instance to select the check box.
Document Automation does not perform post-processing on connected learning instances that contain Python custom logic with either of the following:
  • Single quotes, such as print('Test'). Ensure the Python script uses only double quotes.
  • Characters from non-Latin languages (Unicode characters), such as äöüñй.
Formula validation is not supported in learning instances with the user-defined document type.
When you process documents using the IQ Bot Extraction package, the Python scripts are not applied properly on the form fields for the classic learning instances connected on Document Automation. This results in improper data extraction in the output CSV files and the Validator.
When you click the Table icon on Form OCR Testing Tool (FOTT) editor, the text appearing on the pop-up window is not readable.
If IQ Bot,Control Room, and Bot Agent are installed on the same Windows Server 2012 , then data extraction fails for any Classic learning instances connected in Document Automation.

IQ Bot

What's new
When user logs out, lock released immediately on documents in Validator

When a user logs out of IQ Bot, the lock on the documents in validation is now released immediately so that other users can view or validate the documents. Previously, when a document was in the Validator queue, if the logged out, the lock on the document in the validator was not released immediately, as a result of which other users could not view or validate the document.

Use the IQ Bot Validator

IQ Bot Database Migration Assistant Tool now supports Windows Authentication

When you migrate from Enterprise 11 to Automation 360 by using the Database Migration Assistant Tool, you can now use Windows Authentication for connecting to the IQ Bot database server.

Choose classification type while creating learning instance

When you create a learning instance, you can now choose which aliases and fields are considered for classification by using the Disable classification changes on domain modification (fields, alias) check box.

  • When this check box is selected, the classification process uses only the aliases or fields that were present in the domain when the learning instance was created. Aliases and fields added to the domain after the learning instance was created are not considered for classification.
  • When this check box is cleared, the classification process uses all the aliases or fields that are part of your domain. Aliases and fields added to the domain after the learning instance was created are also taken into consideration for classification.
This feature classifies the document to the same group even if there are any updates in the domain.
What's changed
Upgraded machine learning models

IQ Bot now uses an upgraded machine learning model for detection of check boxes and Advanced tables.

Python libraries upgraded to 64-bit

All the Python libraries provided in IQ Bot by Automation Anywhere for this release are now upgraded to 64-bit from 32-bit. If you are using third-party libraries, ensure that you upgrade to 64-bit libraries. After upgrading the third-party libraries, ensure that these libraries work successfully.

Fixes
When you download an unclassified document, click Delete from the server, and then toggle the learning instance from production to staging and back, the document no longer appears in the untrained folder. Previously, when you downloaded an unclassified document from the untrained groups, even after you deleted it, the document still reappeared in the folder.

Service Cloud case ID: 01781935

When you upload documents, the Classifier no longer halts or becomes stuck. Documents that are not classified are now cleared from the queue, allowing smoother processing of subsequent documents. If these documents fail to be processed within the timeout limit, they are now categorized as unclassified and you can download these unclassified documents.

Service Cloud case ID: 01762844, 01272642, 01750753

You can now install IQ Bot by using Windows authentication and CyberArk Vault and then rotate the key. Previously, when you tried to rotate the Vault key, the following error message was displayed: Error occurred while setting up the new key. Please check the logs
You can now select multiple learning instances for backup by using the Search drop-down list. Previously, after you clicked the Create Backup option on the Migration Management page, if you selected a learning instance and then used the Search drop-down list to select another learning instance, the previously selected learning instance became deselected.
The value displayed in the My totals > Accuracy field now matches the value in My learning instances > Field accuracy for the available learning instances.
You can now add documents to manually created groups and upload documents to staging.

Previously, when you edited a manually created group or tried to upload a document, you encountered an error and the following error message was displayed: Failed to update manual group: Duplicate label found.

Service Cloud case ID: 01849616, 01845590

You can now use the migration utility to successfully migrate learning instances created for a custom domain. When you import IQBA files with duplicate fields (even through an API), the following message is displayed alerting you to the existence of duplicate fields: Error: Duplicate field found Invoice number. Previously, the migration utility could not be used to migrate learning instances created for a custom domain.

Service Cloud case ID: 00729627

During migration, you can now successfully import and export IQBA files. Previously, the import and export of IQBA files failed because the content classification table contained duplicate training fields.

Service Cloud case ID: 00826891

When a document is successfully downloaded, it is now placed in the Success folder. Previously, documents that were successfully processed were sent to the untrained folder, causing inconsistencies in the download process.

Service Cloud case ID: 00780227,00842077

When there is an update in the domain, IQ Bot now classifies a document to the same group.

Previously, whenever there was any update in the domain, IQ Bot created new groups for the same document.

Service Cloud case ID:01622981

If a document without content is uploaded to a learning instance in production, the following message is logged in the bot_launcher.log file: Unable to upload file <filename> to the Learning Instance. No content in file, size 0" whenever there is a 0 KB file is uploaded.
The security vulnerability related to Unquoted Service Path has been fixed. Previously, the Windows host referred to an installed service that used an unquoted service path containing whitespace Automation Anywhere Cognitive Console: D:\Program Files (x86)\Automation Anywhere IQ Bot\Portal\node modules\winser\bin\nssm64.exe, which resulted in the vulnerability.

Service Cloud case ID:01880370

Limitations
You cannot create a bot in IQ Bot when an invalid document is uploaded to a custom or manual group. Also, when an invalid document is uploaded along with a valid document and processed for extraction, the See extraction page keeps loading.
IQ Bot does not support Oracle Database for on-premises installations.
During production processing, if a learning instance is switched from Production to Staging or the other way around, for any documents processed during this switch, the production document that appears in Validator is blurred and not in human-readable form.
IQ Bot does not perform post-processing on a field if the custom logic uses Python script containing single quotes, such as print('Test'). Ensure the Python script uses only double quotes.

Custom Logic Python script challenges in A360.26 IQBot (A-people login required)

If your IQ Bot custom logic uses the Pandas library, the script must have the import Pandas statement.
IQ Bot does not perform post-processing on a field if the custom logic uses Python script containing characters from non-Latin languages (Unicode characters), such as äöüñй.
In IQ Bot, when you run a script created on versions prior to Automation 360 v.26, you might see a warning. However, this does not impact the script execution and the functionality.