Smartops Release Notes

Click here to see this page in full context

UST SmartOps™ uses cognitive automation to intelligently and continuously accelerate business opportunities by reducing costs, streamlining efficient operations, and uncovering unrealised growth revenues for your business.

SmartVision Release 2.5

Release Date: 26 October 2021

Provision to support multiple OCR engines

Architecture Upgrades

Detailed error messages for projects that are not ready

Introduction of "Transfer learning of models"

Train through document annotation (as-is process)
Re-train an existing global model
Re-train a custom model

Spanish Invoice document extraction

Re-send escalated documents

Provision to rotate document image

Flexibility to approve documents with either Insights or Line Items

Updates to field level Accuracy calculations

Provision to add soft validations

Explicit configurations for Auto fill option

Overview

In 2.5 release, DU will start supporting multiple OCR engines like OCRmyPDF and Azure OCR. In addition, the DU architecture is revamped to build a scalable architecture that support timely customer onboarding and ensure collaborative development.

DU Trainer is upgraded to support "Transfer learning of models". User will be able to re-train an existing model for additional FOIs, alter any existing FOI [Add/Remove fields]. This feature is applicable for Insights & Table FOI training. Three options will be provided to the user for retraining: train from base model, train from global model and train using custom model.

Upgrades in SmartVision PWF include, provision to rotate document image for incorrectly oriented documents. This would enable the proper usage of point & select and table re-bounding features. Provision to add the comments and re-send escalated invoices to the originator, field level accuracy calculation updates, provision to configure soft validations for informational purpose, enable explicit configurations for Auto fill option, provision to approve documents with either insights or table data.

The release also includes a pre-trained model to support extraction from Spanish invoice documents.

Provision to support multiple OCR engines

In DU, users will have the provision to choose any one of the available OCR engines based on customer needs. Default OCR engine is Tesseract and other options include OCRmyPDF and Azure OCR. The configuration for additional OCR engines needs to be done during deployment.

Architecture Upgrade

The scalable DU architecture broadly includes 4 layers – AI components, DU , PWF and SmartVision

AI Components - Train, test and infer AI components. Includes Classification, Extraction, OCR, Image/Document Enrichment/Correction
DU - Reduce pipeline to allow DU to get insights for any specific document type. Seamless integration with AI Components and global model library
PWF - Document Ingestion, document enrichment and classifications to be handled by the workflow. Supports customer specific configurations, pre-processors and post-processors. Allows seamless integration with AI components and DU and pre-built insights workflow
SmartVision - Create a business solution using a combination of workflows. E.g. Claims Settlement, Loan Processing, Exploratory Search, Document Linking etc.

This will enable timely customer onboarding as well as allow other teams to contribute post processors, classifiers, extractors etc.

Show detailed error messages for projects which are not ready

Going forward, DU will display the following messages (in detail) when the project is not ready.

Document Type Classifier is not selected
Field configuration is not complete
Incomplete Project Configuration

Transfer Learning of Models

DU trainer is upgraded to support "Transfer learning of models". This would allow users to re-train an existing model for additional FOIs or alter any existing FOI [Add/Remove fields]. Feature supports both Insights & table models. DU Trainer supports three options for re-training.

• Provision to train through document annotation (as-is process)
Field level accuracy graph will include adjusted accuracy in addition to accuracy displayed today. This calculation would primarily exclude the formatting applied on extracted data to meet the target requirements. For e.g.: if 1234AX is the extracted data and approved data is 1234 (AX is removed), it would be considered as match.
• Provision to re-train from global model
In addition to document annotations, user has the provision to choose the global model on top of which the training needs to be done. Model training uses the global model picked by the user as the base and applies the annotations on top of it.
• Provision to re-train from custom model
In addition to document annotations, user has the provision to choose a custom model on top of which the training needs to be done. Custom model can be either chosen from the region or uploaded by the user. Model training uses the custom model picked by the user as the base and applies the annotations on top of it.

The model selection is allowed only for initial training. Second iteration onwards, the latest model available in the project will be picked as the base for training. To re-use the fields from the base model, user will have to include those fields as part of annotation.

Ability to re-use files & annotations across DUT projects, Model evaluation and Assisted Continual learning is planned in upcoming releases.

Spanish Invoice document extraction

Pre-trained SE model to support data extraction from Spanish invoices. DU will then support Spanish invoice documents.

Resend Escalated invoices

Going forward, AP Supervisor will have the provision to re-send an escalated invoice back to the originator after making required changes or with a comment. Changes made by the supervisor will be available for the AP clerk when he receives the document. This allows all easy tracking of clarifications and also improves the bandwidth of AP Supervisor.

Provision to rotate document image

In document review screen, image view will include the rotate provision. The purpose of this feature is to enable users to use "Point & Select" & "Table Rebounding" features for documents which has incorrect orientation today.

Feature Enhancements in SV PWF

• Field level accuracy calculation updates
Field level accuracy graph will include a new graph for Insights and will be named Insights Accuracy which will be the default graph. This graph will exclude the formatting applied on extracted data to meet the target requirements. For e.g.: if 1234AX is the extracted data and approved data is 1234 (AX is removed), it would be considered as match. Existing Insights Accuracy graph would be renamed as Insights - Raw Accuracy.
• Provision to add soft validations
User will have a provision to define soft validations in addition to hard validations available today. These are warning messages and can be overridden by the user. Applicable to internal & external validations messages.
• Enable explicit configuration for Autofill
During project creation, the autofill configurations will be explicitly captured in the external configuration section. This would eliminate the dependency on custom layer to maintain dummy configurations as part of customer deployments.
• Ability to approve documents with Insights or Line Items only
User will have the provision to approve documents with Insights or Line Items.
• Ability to reuse OCR text for downstream processing
• Dynamically adjust preview screen when review pane is enabled

(4220) Variable created not listing in drop down in workflows - Pre prod
(4264) Navistar: Point and select value copied was retained when edit to second field.
(4452) RPFC AP Production: failure in processing invoice in DU
(4556) Navistar : Split was not happening properly
(4585) RPFC AP: No date invoice was loaded with current date
(4644) RPFC AP Production: Invoice missing in IE, but success in DU
(4794) RPFC AP Production: IndexError
(4845) RPFC AP Production: Customer Number Accuracy is less
(4846) RPFC AP Production: Invoice GST Accuracy is less
(4847) RPFC AP Production: Shipping Address Accuracy is less
(4848) RPFC AP Production: Supplier ABN Accuracy is less
(4849) RPFC AP Production: Vendor Name Accuracy is less
(4994) Navistar du extraction failed due to version conflict (du.batch.inv.page_split)
(5037) RPFC AR Production Cluster 2- Validation URL not shown in UI
(5161) DU Preprod : elasticsearch version_conflict_engine_exception
(5182) Navistar: Batch list not getting refreshed after removing the date filter.
(5221) Navistar: Save & Approve button had UI issues
(5283) IQVIA-SV Review screen pdf area is blank when using Vespa for extraction
(5359) RPFC AP Production: ConflictError
(5383) Navistar: System Failure_ConflictError
(5454) RPFC AP Production : SYS_FAILURE - Celery Task - failed - ModuleNotFoundError
(5491) RPFC POP : DU Error Unexpected system failure during step du.batch.inv.page_split
(5545) Navistar : Send batch count is mismatched in Advanced tab.