Vantage 2.3.2

Vantage Application #: 0.47.0

Platform #: 0.30.2

Technology Core #: 2.42.0

Vantage Solution #: 0.24.1-cloud

Release date: 05.08.2023

Updates to existing built-in skills


  • Invoices JP: extraction of official company IDs is now supported.
  • Invoices UA: new skill for processing Ukrainian invoices.

Purchase Orders

  • Improved field extraction quality after training the neural network on more documents.

Brokerage Statement

  • Improved field extraction quality after training the neural network on more documents.


  • Extraction of line items is now supported. Extracted columns: Description, Quantity, Price, SKU, Amount.


  • Processing of handwritten text is enabled for forms:
  • 1095-C
  • 1095-B
  • Address parsing is enabled for forms:
  • 1099 Composite
  • W2-G
  • 1099-OID
  • 1099-G
  • 1099-MISC
  • 1099-NEC
  • 1099-INT
  • 1099-B
  • 1099-R
  • 1099-DIV
  • 1098

New built-in skills

The following new skills have been added in version 2.3.2:

  • Loan Origination skills
  • Initial Escrow
  • Account Disclosure Statement
  • Desktop Underwriting
  • Appraisal 1025
  • Appraisal 1073
  • Deed of Trust
  • TaxForms
  • 1095-A, Health Insurance Marketplace Statement
  • W-2G, Certain Gambling Winnings
  • 1099-OID, Original Issue Discount

Classification skill

  • Online Learning for a Classification skill with multi-document transactions


Vantage 2.3.2 supports the following four languages for ICR:

  • English
  • German 
  • French
  • Spanish

Improved ICR recognition on fields with combinations of letters and digits for English

Improvements in OCR technologies

Further improvements in the quality of Japanese OCR

The quality of Japanese OCR in Vantage has been further improved for condensed and expanded characters, as well as half-width katakana, with fewer OCR recognition errors caused by diacritics.

Rotate incorrectly oriented pages in digital-born PDFs

Mobile Input

For detailed instructions on working with Vantage Mobile Input, see:

Special interface for capturing passports

Documentation on Vantage Mobile Input internet communications

Administrating Vantage

  • Server-to-server authentication scheme now supported

ABBYY Vantage Advanced Designer

  • Added ability to search for two related companies
  • Testing Deep Learning results after each epoch
  • NLP Language support

Starting from version 2.3.2, Vantage offers full NLP support* for:

  • English
  • French
  • Spanish
  • German
  • Russian
  • Japanese (preview) 
  • Portuguese
  • Italian
  • Dutch

*”Full NLP support” means support for Segmentation, Named Entities (NER), and NLP Deep Learning activities.

In Vantage 2.3.2, the following bugs have been fixed:

  • In the "Select Skill" dialog box, skills were not sorted alphabetically.
  • The extraction quality of Line Items was improved for the Purchase Orders skill.
  • Pages were not rotated correctly.
  • A “400 Bad Request. Request Header or Cookie Large” error occurred when trying to authenticate via Azure Active Directory.
  • Dots were not recognized in decimal values.
  • Extraction slowed down when a document skill was trained on a form with a lot of text.
  • Trying to log in to Vantage in Incognito mode or after clearing the browser cache resulted in the cursor spinning indefinitely.
  • Extraction training on a document set containing 10000 invoices failed with the "Could not set vantage model, because could not find training step" error.
  • Some Process Skills were not shown in Skill Monitor after being published.
  • A "Get Results: Header 'Content-Length' is missing or invalid." error occurred while using UiPath connector.
  • Training Invoice EU in full text recognition caused horizontal fields to incorrectly be displayed as vertical fields.
  • Extra spaces between numbers were detected on some documents.
  • The recognition quality of line items was low when processing Chinese invoices.
  • Training for new columns succeeded in Advanced Designer but failed in Vantage.
  • Original file names (containing symbols) was not preserved on import.
  • The paragraph separator symbol was detected when performing full text recognition, causing problems with Catalogue lookup.
  • The slash symbol (/) was not recognized correctly in full text recognition.
  • Unknown symbols in the recognition results for some documents.
  • Handwritten numbers were not recognized correctly.
  • A loss of quality occurred when exporting to PDF/A using OCR skill.
  • Extraction field training failed for customer’s skill because of a time-out.
  • Training failed for a specific Document Skill.
  • A copy of Invoice CN skill (v13) stopped capturing fields after it was opened in Advanced Designer and published again.
  • A "Not enough memory" error occurred when restoring a skill from a local folder in Advanced Designer.
  • An empty screen appeared in the Document skill editor for a specific skill.
  • Having a document with a content restriction in the training document set caused the training of Deep Learning for NLP activity to enter an endless loop.
  • Data Catalog lookup rules were not populating data.
  • An "Internal program error "PerspectiveFixerConverter.cpp, 162" error occurred during training.
  • An option to resize columns width was added to Skill Catalog, Skill Monitor, and Document Repository.
  • Records for some steps were not exported via Vantage Reporting API.
  • Language settings were not displayed correctly in Invoice EU skill.
  • Digital-born PDFs with an incorrect text layer could not be rotated.
  • Document Skills in the Extract activity list were not sorted in alphabetical order.
  • The ResultClassName and ClassName tags were missing from the output JSON file.
  • An invalid workflow could be published in Process skill editor.
  • An error occurred in some cases after selecting the data catalog record manually on Manual Review.
  • FlexiLayout Studio projects with the Object collection element were imported incorrectly to Advanced Designer.
  • The "Due Date" field was recognized incorrectly when recognizing a JP invoice document.
  • Max error percentage property was imported incorrectly from FlexiLayout projects to Advanced Designer if its value was set to 0, causing an error and elements to not match.
  • A "RunTime error" occurred when trying to view rule errors on Manual Review.
  • Skill versions that had active transactions could be automatically deleted if the "Keep 3 latest and versions currently in use" option was selected in the skill versions settings.
  • Added the following error message: "Cannot import skills that already exist in the tenant.".
  • ID Skill transaction failed when trying to process non-ID documents.
  • User documentation inaccuracies were found.
  • Some transactions in Skill Monitor displayed an incorrect status.
  • It was not possible to create or delete fields in some skills.
  • Japanese characters were not recognized in line items.
  • Export skill operations failed to create an archive and were stuck in the Active state.
  • Verification results were discarded for non-table fields with regions after applying the continue line items option.
  • Incorrect skill description in the invoice guide on Marketplace.
  • Uploaded password-protected document remained in the set.
  • Classification Skill returned low-confidence results after upgrading to v. 2.3.2.
  • If a copy of a built-in skill was retrained using certain document samples, the quality of the processing of other documents deteriorated considerably.
  • Redundant underscores were added to recognized text when processing documents via OCR skill.
  • Chinese characters typed in any field were not saved during Manual review.
  • The "Column settings" dialog box did not open when choosing the "Column settings" menu item.
  • When using the "Continue table from this row" option, the values in the line items column were displayed in an incorrect format.
  • Filed values were extracted incorrectly for some documents.
  • Redundant characters appeared in recognized text after performing OCR.
  • Vantage automatically logged out users if IP restriction was enabled, even if their IPs were not black-listed.
  • Ability to select all/deselect all barcode types has been added in the "Barcode Settings" dialog box.
  • Some values were not extracted from line items when processing documents using the Document skill.
  • Some numbers were not recognized on certain documents.
  • If, after opening the table with line items for the first document, the operator opened another document in the batch, the values of one column could be moved to another column. This resulted in the values being in the wrong column in the export file.
  • If a whole table was removed and then recreated during Manual Review clicking the “+” button inserted the second row above the first.
  • A wrong Vantage version was displayed in the About dialog box.
  • Skill export could fail if there was a large number of documents in some document set and multiple document sets were queried at once.


Have more questions? Submit a request



Please sign in to leave a comment.