Version 11.3.4 Release Notes for IQ Bot

The release notes for IQ Bot Version 11.3.4 describes new capabilities, changed features, fixed features (resolved issues), security fixes, deprecated features, and known behavior or limitations.

New features

IQ Bot Version 11.3.4 new features
Feature Description
(Beta) Enabled data capture and OCR using Google Vision API for Asian languages

Use the Beta version of the Google Vision API with IQ Bot to improve the accuracy of the optical character recognition (OCR) results for training documents in Asian languages, particularly in Japanese and Korean. Google Vision API files are installed on your machine automatically during IQ Bot installation.

IQ Bot 11.x: Use Google Vision API OCR engine

Magnetic ink character recognition (MICR) extraction

Use the MICR OCR feature with ABBYY FineReader Engine v12.2 to extract MICR data from financial checks.

IQ Bot 11.x: Extract data using magnetic ink character recognition

Option to turn off PDFBox as OCR engine

Users have the option to turn off the PDFBox value to false to use the extraction functionality of the selected OCR engine such as Tesseract4 OCR, ABBYY FineReader Engine, and Microsoft Azure API. Disabling the PDFBox OCR optimizes extraction from hybrid documents that have a mix of images and digital values.

IQ Bot 11.x: Disable PDFBox option

Changed features

IQ Bot Version 11.3.4 changed features
Service Cloud ticket number Description
Enterprise Control Room > Administration > Settings IQ Bot When a user tries to register IQ Bot with the Control Room, the system now shows an error message if you enter an incorrect URL.
UI auto-refresh in IQ Bot is disabled The system does not automatically refresh (auto-refresh) the IQ Bot UI, unless the user loads a screen or clicks Refresh to manually refresh data on a page. This feature helps reduce the response time for loading new data or performing any action in IQ Bot.

Refresh data in Learning Instance details page

- - Previously, the user had to select the certificate from a provided list. Now, during the installation process:
  • The CA.crt is automatically created in the <installation path>/Portal/keys folder.
  • The public certificate is added to the Java keystore in <installation path>\JRE\zulu8.40.0.20-sa-fx-jre8.0.222-win_x64\lib\security\cacerts.
- - No updates are required in the file for IQ Bot Version 11.3.4 to work with earlier versions.

IQ Bot version compatibility matrix

Fixed features

IQ Bot Version 11.3.4 fixed features
Service Cloud ticket number Description
- - After correcting the field values in the Validator when a user saves the changes, the Validator updates only the edited values. The IQ Bot Dashboard now displays the correct accuracy value for that learning instance.
- - After you correct a field value, the ML suggestion works as expected and provides suggestions for that field value in other documents as well. The user is able to select the ML suggested value on the first click.
- - The system now generates a valid response for the AAE_IQ Bot Admin role when accessing dashboard accuracy details for a specific project ID.
00417631 / 00423076 Uploading a document with a long file name does not return an empty learning instance success folder anymore. Instead, the file name is now visible and is successfully processed to the success folder.
- - HTTP response header settings related to security vulnerability are now fixed.
- - A user can now successfully install a later version of IQ Bot even if an existing Version 11.3.3 is installed on their machine. The system does not create duplicate data in the contentclassification table. In case duplicate data is created, the system displays an error message and rolls back the installation. The user then has to manually verify and remove the duplicate data from the contentclassification table.
00358689 The user can no longer see the system-created temporary files after the processing of documents in the IQ Bot server.

Security fixes

Critical security fix: TLS 1.2 is required for HTTPS connections to IQ Bot
To maintain security standards and promote the safety of data, the earlier TLS 1.1 encryption protocol is disabled in IQ Bot. Starting with Version 11.3.4, all inbound connections to IQ Bot must use TLS 1.2.

Users must verify that their browsers are compliant with TLS 1.2.

Deprecated features

There are no deprecated features listed for this version.

Known limitations

IQ Bot Version 11.3.4 known limitations
Service Cloud ticket number Description
- - If the value of PDFBoxOCREnabled is set to False in the settings.txt file, some multilingual documents do not get classified.
00356534 IQ Bot does not support migration of learning instances created using custom domains, unless the custom domains were imported in the same order, and the field IDs between the source and target system match.
- - With Google Vision API OCR, for accurate value extraction, use training documents that are single language only.

For example: Documents cannot have a mix of Japanese and English language fields. They must be fully Japanese only.

- - The system is unable to generate CSV files with the appropriate extraction results for documents with more than 10 to 15 pages that are sent for validation.
- -
The Bot training page displays the bot ID in the URL. If a user manually changes the bot ID in the URL, the system creates a new bot with that ID instead of showing an error message.
Note: IQ Bot does not support manual modification of the bot ID in the URL.
- -

Sometimes the user is unable to edit or add new fields to existing learning instances in a cluster environment. The new field is not available in the Designer for training, as it did not get added to the learning instance. In such cases, reach out to support.

This does not affect learning instances in the standalone mode.

- - The Validator does not stop data extraction at the end of a table/repeated section for the first table, but continues on to extract data from different fields in the following tables as well. This leads to incorrect data extraction and the result is shown in the See extraction results screen for the first table.
Tip: As a workaround, we recommend using the Advanced table options setting in the Validator and enabling the Stop extraction at "End of table indicator" check box. This stops the table extraction at the table footer.
- - When a user tries to install a newer version of IQ Bot, the system uninstalls any existing previous version of the product on the user machine. The system then installs the newer version without displaying any notification for the user.
- - When a user with an AAE_IQ Bot Admin role tries to create a learning instance with IQ Bot extensions, the system displays a forbidden access with admin role message.
Tip: As a workaround, we recommend assigning the user with the AAE_Basic role in addition to the AAE_IQ Bot Admin role .
- - The image invoice document classification is inconsistent. The system deletes such documents from the temporary folder after an upload, which leads to results showing that no classification occurred.
- - If a user has configured a new output and log path in the Settings file, the system fails to update the change. This results in the audit logs being unavailable at the new location.
- - The data validation error checklist in the system is not working according to defined priority, thus impacting validation results.
- - IQ Bot is unable to generate system identified regions (SIRs) for PDF files with multiple formats in different pages.
Tip: As a workaround, we recommend turning off the PDFBox option as an OCR engine. IQ Bot 11.x: Disable PDFBox option
- - In the Learning Instance details page when a user deletes a learning instance, and then navigates to the Bots tab and sets the same learning instance to production, the system fails to display an error message stating that the learning instance cannot be found.
- - IQ Bot is unable to identify some check boxes, or is extracting incorrect values from some of them from the uploaded documents.
- - The IQ Bot classifier performance is inconsistent for the same learning instances, and for learning instances with the same domain and custom fields.