Document Automation and IQ Bot v.29 release

Download as PDF

Feedback

Updated: 2024/05/17

Review what's new and changed, and the fixes and limitations in IQ Bot and Document Automation for the v.29 release.

Important: Starting from this release, we have moved information about updates in each package (such as what's new, changed, fixes, and limitations) to package-specific pages. You can access these pages by navigating to Package updates overview > <package-name> releases.

For a consolidated list of packages updated in this release, see Packages updated in v.29.

Document Automation

Search:


What's new

Validate table data quicker using auto fill You can use the Auto fill button in the validation interface to extract hundreds of rows of data from multipage tables. After you manually validate three cells in a column, auto fill extracts the remaining values from that column in the document.
Use Advanced Classifier package in Document Automation You can use the Document Automation Advanced Classifier package to classify documents and pages, and split documents into a folder structure. You can upload documents from the folders to their respective learning instance to extract content in Document Automation. You must set up a Skillja license in a credential vault to run a bot containing actions from this package. Advanced Classifier package
Document Automation support for Google Custom Document Extractor You can now create a user-trained learning instance in Document Automation and extract it using a Google Custom Document Extractor (CDE) processor. You can use this new capability to train a model using Google CDE for any document type, and it supports 30 languages. After you deploy a model, you can embed the processor URL within the Document Automation extraction process. To use this capability, you must have an Automation Anywhere Document Workspace license and a valid subscription to Google CDE.
Improved heuristic feedback with form and table extraction Document Automation provides an improved extraction through heuristic feedback with a focus on complex scenarios, such as multi-tables. Additionally, there are extraction improvements for both form fields and out-of-the-box performance (specifically for table fields).
Document-level data validation in Document Automation With document-level data validation, you can create rules to ensure the accuracy of extracted data across multiple fields in your documents. You can define various conditions, such as pattern matching or equality checks. When those conditions are met, you can effortlessly take action to flag errors or warnings, clean up or replace values, or set new values. Validation rules in Document Automation

Search:


What's changed

In Standard Forms, you can create and train new models in Azure AI Document Intelligence v3.0 only. You can use existing models in Azure AI Document Intelligence v2.1, however, you cannot create new models.
Azure AI Document Intelligence Studio integrated with Document Automation is compatible only with v.3.0 model for all future model creations.
As of Automation 360 v.29, Document Automation Standard Forms supports the check box type for forms and table fields. The Standard Forms learning instance has an additional check box field data type and the extraction result can be one of the following: Checked Unchecked Not found (if the system is unable to extract the result) For any Standard Forms learning instance that was created before Automation 360 v.29 and has a check box field, you must update the data type for the check box field at the learning instance level.
Enhancements were made in the Validator to improve the validation process in Document Automation.
For Automation 360 v.29 release and later releases, the learning instance export and import functionality has been enhanced to include the ability to copy heuristic feedback data with the configuration. You can seamlessly move your feedback data from one environment to another without any data loss. As a result, this feature significantly reduces the need for rework, ultimately saving valuable time

Build 18345: This build includes the following fixes (along with fixes from the previous builds):


Fixes
In Validator, if you enter a valid entry or data in the field, and click outside of it, the field is validated. This was a limitation in previous builds 18332, 18302, 18277 (Cloud-Sandbox) and build 18324 (On-Premises).
When Extract field using pattern is chosen in the learning instance and the document is processed, the field is validated on entering valid data. Here, only the respective field is affected, and its extraction is based on the provided regular expression input. This was a limitation in previous builds 18332, 18302, 18277 (Cloud-Sandbox) and build 18324 (On-Premises).

Build 18277: This build includes the following fixes:

Search:


Fixes

If a table name in the Validator has more than 50 characters, the field name in the table is now displayed when you hover over it. Previously, the table name was not displayed.
If a field name in the Validator has more than 100 characters, the scroll bar appears and you can see the value of the form field. Previously, the form field name was not displayed.
The run analysis result is displayed in Standard Forms when creating a v3.0 model with multiple pages.
You can now use the advanced check-out options for bots that contain actions from the Document Extraction package.
If you upload documents to a custom process instead of a Document Automation auto-created process, the Download bot now successfully downloads the CSV of extracted data and deletes the uploaded files from temporary storage.
The Automation Anywhere Control Room now tracks Document Automation license consumption in newly-installed cluster Linux environments.
Imported learning instances now have pattern-based validation enabled by default.
You can now submit a document that has a pattern validation rule with a regular expression that contains the pipe symbol (\|).
In the Bot Insight, the Unknown option under OCR Engine attribute for Google Document AI provider has been removed from the Document Workspace dashboard. Now the filter displays ABBYY as the OCR engine, ensuring better clarity and accuracy in filtering options.
Extraction is now successful when creating a learning instance in Document Automation using a custom domain. Previously, uploading a document for the learning instance using a custom domain would result in extraction failure, but this issue has been resolved. Service Cloud case ID:01967442
Documents with Windows OS user-names containing more than 17 characters can now be processed successfully in the latest package of Document Automation. Previously, attempting to process such documents would trigger an error message stating, The File cannot be processed, it may be corrupted. Please contact technical support if the file is valid. Service Cloud case ID:01951382, 01950245
The Document Extraction package now successfully extracts documents when the option to store the output files on the server is chosen in a stand-alone setup. Previously, selecting the option to upload the output to the server would result in a failure of the Extraction action, accompanied by the error message DW_EXTRACT_FAILURE 500. Service Cloud case ID:01943058, 01972932

Search:


Limitations

The check box feature (in preview mode) might result in inconsistent extraction for the check boxes field, which could lead to inconsistent results. In such cases, if the system is unable to accurately extract the check box field value, it will be labeled as Not Found.
If a learning instance name contains a space, then the downloaded file name will have the space replaced with the encoded value %20. For example, if the exported learning instance name is `Test 123`, then the file name is displayed as `Test%20123`.
With the implementation of new document level rules, if the condition is based on a Form Field and the resulting action should affect a Table Field, the expected outcome is not reflected in the Validator.
You will not be able to use the Extract field using pattern for a table field with a custom alias. Workaround: On removing the regular expressions (regex), the extraction works as expected with the custom alias.
When the table name in the Azure AI Document Intelligence Studio UI exceeds 99 characters, the UI does not handle it properly.
On creating a validation rule with does match regex condition and show error action, a custom error is shown for a field with no extraction. Workaround: If condition is does match regex, add another AND along with is not empty condition to use the action show error. This will not display any error over fields with no extraction.
While creating or editing a learning instance, if you select the Extract field using pattern check box and then subsequently clear it without entering a value in the regular expression field, the Create button remains disabled even if all other mandatory fields have been entered. Workaround: Enter value in the regular expression field, and then clear the Extract field using pattern check box.
If you select the Chinese language in Process Composer, the auto-fill button label does not update to reflect the selected language during validation.
If you process a document with Google Custom Document Extractor, the extraction is failing for the below languages: Albanian Croatian Finnish Icelandic Slovenian Vietnamese
(Service Cloud Case ID: 01981332, 01981088) During extraction, the heuristic feedback does not provide values for certain fields in cases where the previous mappings contain non-English characters, such as Arabic characters or text with single quotes (').
While editing a learning instance, if you use the Up or Down buttons to change the value of the Confidence level and click the Update button, the changes are not retained. Workaround: Ensure that you enter the value that you want to increase or decrease for the Confidence level.
For the Community Edition: Google Vision OCR will not be displayed as it is not supported. Document extraction using Google CDE or Google Document AI (User-defined) is not supported.

IQ Bot

Search:


Fixes

The first time you click a learning instance connected from Automation 360 IQ Bot, you are directed to the details page for that learning instance. Previously, you were redirected to the IQ Bot home page instead.
Automation 360 IQ Bot now shows the correct error message if you upload a corrupted document to a manual group.
You can now use more than 54 categories for the Train and Retrain actions. This allows classification according to the provided training model files. Previously, the Classifier was unable to classify the files into categories if the number exceeded 54 categories.
You can now create and run the Document Classifier package with a variable in the child task bot. Previously, when you ran the task bot command with a variable to call the child task bot, it displayed an error.
You can now view the full Python script that is added to the Python Logic tab when you navigate from fullscreen to smallscreen after a test run. All the options such as Test Run, Clear Box, fullscreen, and Save are also visible. Previously, when you navigated from fullscreen to smallscreen, you could not view the Python script and the other options were hidden.
Healthcheck for an IQ Bot instance shows the correct OK status even if the instance is idle or accessed after a day. Previously, healthcheck showed a status of NOT_OK when the IQ Bot instance was idle or accessed after one day.
You can now export, import, upload, and download an IQ Bot archive (IQBA) file with a maximum file size of 5 GB. Previously, if the IQBA file size was greater than 268 MB, then the file storage did not upload the exported file from the output folder to the desired location.
You can now import an IQBA file successfully into the learning instance. Previously, IQBA import was failing when the exported learning instance had certain training embedded in it with line break characters.

Updates to the interface

Review the updates to the interface made to the interface in Document Automation for Automation 360 v.29 release.

<span> Video demonstrates changes to the interface in Document Automation </span>


With document-level data validation, you can create rules to ensure the accuracy of extracted data across multiple fields in your documents. Document-level rules for improved extraction
You can leverage export and import functionality with the ability to copy heuristic feedback data with the configuration. Validation feedback in export and import of learning instances
Create a user-trained learning instance in Document Automation and extract it using a Google Custom Document Extractor (CDE) processor. Support for Google Custom Document Extractor (CDE)
You can create and train new models using Azure AI Document Intelligence v3.0 only. Standard Forms support for v3.0 model