Document Automation v.32 release

Review what's new and changed, and the fixes and limitations in Document Automation for the v.32 release.

Document Automation

What's new
Manage validation feedback using lock - unlock feedback feature

You can now lock the validation feedback for a learning instance using the Lock - Unlock feedback feature. With this feature enabled, you cannot provide any further validation feedback to respective learning instance.

Note: You can use this feature only when the Improve accuracy using validation option is enabled.

Lock the validation feedback
Document Automation Cloud capabilities for On-Premises deployment
You can now leverage Cloud capabilities for Automation 360 On-Premises deployment. With this feature, you can use the following features for On-Premises deployment:
  • Search query using generative AI
  • Google Document AI
  • Google Vision OCR
  • Bill of Lading, Waybill, Arrival Notice, Packing List document types

Enable generative AI and other external connections to Document Automation
Improved user experience through multi-table support in a learning instance

As a Document Automation user, you can now create multiple custom tables in a learning instance. With this feature, you can add custom tables while creating and editing a learning instance for all document types.

Create a learning instance in Document Automation | Create learning instance with generative AI for unstructured documents
Data extraction enhancement with GenAI for table fields
We have extended data extraction with GenAI support for table fields as well. This capability can be availed for unstructured and semi-structured documents from this release. You can extract table data efficiently using simple natural language queries, making document processing faster and more accurate.
Note: This feature is available on Cloud and On-Premises.

Document Automation data extraction using generative AI
Bring your own license (BYOL) support for MS OpenAI

We offer the BYOL option to our users to allow them to retain access and control of their data by using their own account while continuing to use the features and capability of Document Automation.

When you use the Document Extraction > Extract data action in a Task Bot, you can extract data from documents by using Google Document AI or MS OpenAI services.

You already have the option to use your own license key when using the Google Document AI services. From this release, we bring you the option to use your own license and credentials for using MS OpenAI services.

For MS OpenAI services, you would provide the Endpoint URLs and the Service accounts for the GPT and Embedding models, for connecting to Document Automation.

Extract data action | Document Automation data extraction using generative AI
What's changed
Korean languages support for Standard Forms

We now support Korean language along with associated locale for the Standard Forms. This enhancement enables you to process the documents and extract data in Korean language.

Languages supported in Document Automation
Fixes
On the Learning instances page, the Google Document AI provider icon is now visible for Google CDE learning instances.
When a user applies a formula validation across multiple columns in table to validate a value at form field level (Example, Sub Total) and if the user deletes one or more rows and resets the value of form field to match with the present rows, the validation error no longer exists. Also, no error message is displayed.
With improved table extraction for Google OCR, user can no longer view the extra symbol after extraction and the document is now extracted correctly.
When you extract a document that contains table data, the data is extracted correctly with enhanced table extraction.

Previously, in such cases, users were facing table data extraction issue.
When a user connects and processes a document with IQ Bot learning instance that contains non-English characters in document type (such as Arabic and Chinese characters) and is connected to Document Automation, the license consumption is now tracked in the Control Room > IQBot Pages for such connected learning instances.
When a user deletes a parser from Parser configuration page and if a user creates a learning instance later, the Document type drop-down no longer shows the domain associated with the deleted parser.
When you configure a parser and select the Japanese language, the Locale field shows the associated and correct locale (Japanese (Japan)) value.

Previously, in such cases, the Locale field showed the English (United States) value.
As per the required use case, users can now move the field rules up and down while creating a learning instance, and change the field rule order successfully.

Service Cloud Case ID: 01996145
With improved logic for the Update document data action, users can now view all the table data provided through multiple rows in the input JSON file. It is applicable to DocumentJson and DictionaryType document data inputs.

Previously, in such cases, data only from the first row was populated in the table.
The text within the buttons no longer appear compressed in Community Edition. Also, the text padding is now consistent for all the buttons.

Previously, compressed appearance and inconsistent text padding led to bad user experience.
The SIR generation no longer changes after every feedback and you now get the consistent extraction results.

Service Cloud Case ID: 02107121
Limitations
When user processes a document to a learning instance that is imported into Document Automation using IQ Bot - DA Bridge package from classic IQ Bot, the extraction fails if the classic learning instance is created with check box fields and the Improve accuracy using validation option is enabled.
When user processes a document on custom process learning instance, the validate document count is not updated post extraction. Also, if user submits the document, the validate document count is updated with the negative value.
When you process a document with a learning instance that contains multi-tables, the validation feedback does not work for any of the multi-tables.
For bridged learning instances, if training is not provided in IQ Bot or table column headings are not mapped to column fields in Document Automation correctly, the validation feedback is not working for default or custom table fields.
Workaround: To fix this issue, perform one of the following steps:
  • Provide validation feedback using the Advanced training settings and map the header for all the empty columns. The validation feedback will be applicable for next subsequent documents.
  • Provide validation feedback when processing the document for second time and submit the document. The validation feedback will be applicable for next subsequent documents.
Limitations from previous releases

When you process a document using Google CDE with bring your own key (BYOK) setup and the corresponding processor is using foundational model, the document processing fails due to transformational failure.

Workaround: To fix this issue, use the custom model instead of foundational model in Google console.
For a public process, you might encounter an error message in the following scenarios:
  • After validating all the documents in the validation queue.
  • After processing some documents with same learning instance, if you open the first document and click Refresh.
Note: No error message is displayed for the private process.
No error message is displayed to an user when the IQ Bot learning instance has already been bridged by another user.
When you process a document and if it is sent to the validator, you might encounter an issue with decimal numbers (such as .78, .99) for the number data type.

Workaround: To fix this issue, you must enter the decimal numbers as 0.78 or 0.99.