Description

ABBYY Vantage is an Intelligent Document Processing platform providing easily consumable AI “skills” that understand your documents in a fast and simple way. Vantage makes today’s digital worker, and processes, smarter with pre-trained skills that read, understand, and extract insights from documents, helping enterprises accelerate digital transformation.

 

 

Features & Benefits

Vantage skills enable intelligent automation platforms, digital workers, and automation robots to intelligently process business documents in a human-like manner—to understand, extract meaning, and make decisions.

A skill offers a no-code design approach for citizen developers to train document models and automate document processes for all types of documents.

• Core AI skills to process any document

The Vantage platform comes with a set of core AI skills that provide the foundation for processing documents of any kind—structured, semi-structured, or unstructured, and all type of data including machine printed, hand printed, barcodes, signatures, and check boxes.

• Trained document skills to get off to a quick start

Trained skills can be quickly designed to understand and extract information from all types of documents, or users can browse the ABBYY Marketplace for pre-trained document skills—invoices, purchase orders, receipts, W-2 forms, utility bills, insurance claims—to save development time and gain quicker ROI.

• No-code design of skills

Easy no-code skill designer allows citizen developers to design, train, and publish document skills for all types of structured and unstructured documents.

• Monitor and analyze document processes

Measure and analyze the performance of all your deployed skills—so you can continuously improve and move your document process automation to the next level.

• Integrate with any intelligent automation system

Out-of-the-box integration with intelligent automation platforms including Blue Prism, UiPath, Automation Anywhere, and more makes all Vantage skills auto-discoverable from  RPA, BPM, Chatbots, mobile apps, and more.

• Improve accuracy with continuous learning

Vantage skills are continuously getting smarter and more accurate over time, as new document variations and statistical data is collected during human-in-the-loop review. The impact of this makes your business processes more efficient over time and require less human intervention.

How Vantage works?

Start with a skill from the ABBYY Marketplace, or train and publish skills, making them discoverable to any automation tool—RPA, BPM, ECM, systems of engagement, and mobile devices.

• Design and train a skill

Start by designing and training a skill using the Vantage Skill Designer or select from a catalog of pre-trained skills available in the ABBYY Marketplace.

• Publish your skill

Publish your skills to the Vantage Catalog, making the skills discoverable to any process automation platform or workflow using one of the pre-built connectors. Develop your own integration to other systems using the Vantage REST API.

• Monitor performance

Once skills are deployed, Vantage then monitors, measures, and analyzes performance of all your deployed skills, creating new learning models—so you can continuously improve and move your automation to the next level.

Architecture

Deploy the platform on-premise in a private cloud or use the ABBYY Cloud, and scale as needed to meet all your needs.

Vantage is designed as a cloud-first platform that can be deployed on-premise in a private cloud or accessed in the ABBYY Cloud. Vantage utilizes Docker containers and Kubernetes to orchestrate containers, and includes logging and monitoring providing IT teams with required control over the operations of the system.

Integrate the Vantage platform and skills with virtually any application or device using one of the pre-built connectors or the Vantage REST API.

Vantage availability and deployment

• Integral part of enterprise business processes

Enterprises can add Vantage skills to their business processes to deliver key data insights from documents directly into enterprise software systems such as BPM, RPA, ERP, ECM, chatbots, and more.

• Integrated within specialized enterprise software applications

Software Vendors (ISVs) developing specialized enterprise applications for different industry use cases like order-to-cash, claims processing, or customer onboarding can integrate trained Vantage skills to intelligently identify documents and extract data from documents driving greater value for your customers.

• Connected with intelligent automation systems

Enterprises utilizing RPA, BPM, and other automation platforms to orchestrate their process can easily connect trained Vantage skills, automatically classifying documents, extracting business data, and returning the results to the digital worker or workflow process.

New Features of Vantage 3.0

​Update of Built-In Skill Copies into Derived Skills 

In earlier versions of Vantage, you could create editable copies of built-in skills. In this version, you can update these copies, which will also update the Technology Core, and the copy will be converted into a derived skill. Converting a copy of a skill to a derived skill will allow for easy and seamless updates to any future versions of the base skill. 

New Skills 

• Certificate of Analysis – Document Skill
• Denial – Document Skill
• Form 1099-C, Cancelation of Debt
• Mortgage Note – Document Skill
• Riders – Document Skill

 

Vantage and Advanced Designer User Interface Updates

The Vantage and Advanced Designer user interfaces have been improved to make updating skills easier and more intuitive: 

• A new detailed overview of changes and required actions is displayed after each update, making the update process clearer.
• Ability to change field names and types has been added. These changes will be preserved when updating both skill copies and derived skills.
• The What’s New field has been added to highlight the latest updates for each skill version.
• Process Skill
• Redact Fields Prior to Export
• Configure Files Required for Custom Activity
• Mobile Input
• Shortened Links to Vantage Mobile Input

Monitoring and Reporting 

• Analytics GUI

Business Process Reporting

• Ability to specify a certain time periodIn the Vantage API, when downloading a data report from the Warehouse, you can now specify a certain time period, and only those transaction events that were completed within that period will be downloaded. 
• Information about the number of documents and pagesData reports from the Warehouse now contain information about the number of documents and pages in completed transactions. 
• Ability to download big data reportsThe Vantage API now allows you to collect data asynchronously for data reports. This means that data reports from the Warehouse can be split into several files. To get them all, you will need to send a request for each file. This change is designed to prevent failures caused by exceptionally large data reports. 

Ability to Restart a Failed Transaction

The feature allows Processing Supervisors to restart failed transactions caused by temporary issues such as external system unavailability or accidental processing errors. The restarted transactions should process source files from the beginning while preserving all relevant parameters and documents.

ABBYY Vantage Advanced Designer

Skills can now use LLMs

A new prompt-based activity has been introduced, designed to use external Large Language Models (LLMs) within Document skills.  This activity allows you to work with external LLMs in several ways: 

• You can use document fields, text, and images extracted by other activities in prompts for LLMs. For example, you can upload a financial report to extract specific financial metrics (e.g., EBITDA or net profit), extract the data using an Extraction Rules activity, and then compare it to forecast values using an LLM.
• The output generated by an LLM can be further processed, validated, or corrected using Vantage activities. For example, you can validate extracted data against data catalogs to help mitigate potential hallucinations from an LLM.

Deep learning is now possible with smaller numbers of documents

Training can now be carried out with as few as 10 documents; however, a warning will still be displayed for any number of documents below 500. Predefined weights can be used to help mitigate the impact of a low number of documents; however, to achieve sufficient quality with deep learning, you will still need to have a substantial number of documents for training. ​

Advanced Accuracy Reports in Advanced Designer

The following advanced statistics are now available after skill testing: 

• The skill’s quality in extracting field values: Field Precision, Field Recall, and Field F-measure.
• The skill’s accuracy in detecting field regions: Region Precision, Region Recall, and Region F-measure.
• The number of field instances in the reference labeling.
• The percentage of documents containing the given field.

Key Improvements​

Improvements in the OCR Technologies

The following improvements have been made to the OCR technologies 

• The quality of Chinese and Thai OCR has been improved thanks to new e2e Models that have been integrated into the product. This affects all skills running on Technical Core version 3.0
• Improved table detection in documents such as agreements, invoices, financial documents, and mathematical equations.

Improvements in Data Catalog Lookup Rules

The following improvements have been made to data catalog lookup rules: 

• You can now search for a data catalog record that matches any part of any word in the field value, not by just the first word.
• Fields involved in database lookups can be populated by scripts, activities (such as Script or NLP activity), and rules, as well as manually. Starting with this release, if a field value is changed during manual review, the data catalog lookup rule will run again. If the corresponding record is found, the field will become read-only.

And many more bug fixes! You can learn more about new 3.0 features Here.