General
- The maximum number of files that can be added to a transaction is 1,000.
- If more than one file is uploaded via the Public API into a transaction with an OCR, Document, or Classification skill, the files will be merged and processed as a single document.
- Each imported file should not exceed 2 GB.
- Unattended document processing has only been tested for the following numbers of pages: up to 3,000 pages for an OCR skill, up to 500 pages for a Document skill, and up to 100 pages for a Document and Classification skill if the processing flow has a manual review stage. Documents containing more pages may slow down your work with documents in a Manual Review client.
- This release has only been tested with Google Chrome and Safari. Some issues may be encountered when using Firefox, Opera, and other browsers. These browsers will be fully supported in future releases.
- Automatic skill selection for the files uploaded into Documents currently only works for the following languages: English, French, German, and Spanish.
- Excel files (
*.xls,*.xlsx) can be imported with the following limitations:- If an imported file contains a chart with text (for example, a title), this text won’t be extracted.
- Vantage automatically selects a page size from a set of standard sizes. If the contents of an imported Excel file don’t match any standard size (for example, doesn’t fit in terms of width), it may be displayed incorrectly.
- If an imported file contains slicers with data outside the visible area (for example, items that need to be scrolled to), this data isn’t displayed.
- Vantage ignores the Excel ‘Scale to Fit’ option, which reduces the size of the worksheet to better fit the printed page. If the imported table has too many rows, it will be split across multiple pages.
- Vantage may not preserve the relative positioning of a table and an object outside the table. For example, text within the table may shift to the right in relation to a shape above the text.
Process Skill
- If a Process skill contains multiple Classify/Extract activities, online machine learning will be applied to the last activity.
- Manual Review, Input/Output, Assemble, and For each document activities cannot be wrapped in a For each document activity. Wrapping some of them in a For each document activity will be supported in future releases.
- Each ZIP archive to be imported by ABBYY Vantage from a shared folder should not exceed 1.5 GB. Each file inside a ZIP archive should not exceed 100 MB. A ZIP archive must contain files in supported formats only, otherwise the import of the ZIP archive will fail.
- Custom activity supports only synchronous Vantage API calls. A synchronous call implies that processing in the external system takes only a few seconds and that processing results are sent in the third-party system response almost instantly. Asynchronous methods will be implemented in future releases.
- Process skills support no more than 50 Document skills in one Extract activity.
- When importing from an e-mail, only first-level attachments are processed as transaction documents. If there is an attached e-mail file, its attachments will not be imported as files of the transaction.
- When importing from an e-mail, the “FilePath” document registration parameter will not be specified, while the “FileName” document registration parameter will be specified.
- Image fields cannot be exported to CSV. If there is an image field in the Document skill, the value of such field will always be empty in the export results.
- When specifying skill parameters using the API, you can specify no more than 10 key-value pairs in one transaction, with each key containing up to 128 characters and each value containing up to 256 characters.
OCR Skill and OCR Activity in the Process Skill
- Selecting too many languages (more than 20) may slow down performance and impair the OCR skill.
- The maximum recommended number of pages to be processed with an OCR skill is 3,000 pages per file.
- Objects like stamps and stickers in some cases may be skewed on output PDFs.
Document and Classification Skills
- ABBYY Vantage can work with training sets of up to 10,000 documents.
- When new documents are added into a trained Document or Classification skill, the Results tab will show outdated information until the new documents are labeled and the skill is retrained.
- When comparing training results to reference in the Document skill, the Difference tab will not show the error type and reference value for repeating fields.
- A skill can have no more than 100 parameters.
Document Splitter Skill
- Reference labeling and statistics are not yet supported for the Document Splitter skill.
- It is not possible to revert to a previous version of the Document Splitter skill.
- Documents processed with the Document Splitter skill should not contain more than 1000 pages each.
Online Learning
- Online Learning does not support the following types of fields:
- Image fields
- Repeating groups containing image fields, nested repeating groups, groups inside repeating groups or tables
- Generally, the results of Online Learning for checkmark fields are of a sufficiently high quality. However, sometimes checkmarks may not be detected or trained properly because of either the geometric shape of the checkmark field (e.g., round or square) or the way the user chose to select the checkmark field (e.g., if the user crossed or underlined the field instead of placing a checkmark inside). Detection and training of such checkmarks will be improved in future releases.
- The maximum number of documents is 10,000 for the training set and 1,000 for the testing set.
Fast Learning in Advanced Designer
- Deleting a Fast Learning activity from the skill deletes the database lookup rules as well.
Deep Learning in Advanced Designer
- Deep Learning works for text fields only.
- One Deep Learning activity can be trained for 50 fields, including repeating groups and fields, or for one table of up to 32 columns.
- It is not possible to train nested repeating groups and groups inside repeating groups.
Data Catalogs
- A data catalog cannot be deleted or changed if it is used in any skills, including any versions of a skill. To delete a data catalog that you do not need anymore, you must stop using it in the skill (i.e., by deleting or modifying the corresponding rule) and delete all the previous versions of the corresponding skill. Skill versions can be deleted in skill settings. If you do not want to delete old skill versions but want to change the data catalog, simply create a new data catalog.
- A field value is not auto-populated when a lookup rule is used for one data catalog. To populate the field value, you need to manually click the Lookup icon associated with the respective field, select the value from the lookup results, and then retrain the skill.
Skill Monitor
- Only the latest 10,000 errors can be exported to a CSV file from the Error Log. You can use filters and search for specific lines in the Error Log to reduce the number of errors to be exported.
- No metrics are displayed in Skill Monitor if a skill is deleted and then re-imported with the same identifier.
Business Processing Reporting Warehouse and Quality Analytics Reporting Warehouse
- Some of the Process skill transactions are missing from the list of completed transactions.
- When requesting transactions for a specific date range, transactions for the entire period may be returned.
- When getting a list of transactions, if the offset parameter is equal to the totalitemcount parameter, the endpoint fails.
Advanced Designer
- The Number must be within interval property is disabled for the Number and Money field types in the Manage Fields dialog box. You can set this property in the Reference section of the Fields tab or in Vantage.
- Field settings edited and saved in the Manage Fields dialog box are not synchronized with the reference labeling section. To overcome this limitation, edit the field settings either in Vantage or in the Reference section of the Fields tab in Advanced Designer.
Built-in Skills
- It is not possible to enable additional languages in copies of a built-in skill. Built-in skills are intended to process documents in specific languages. Enabling additional languages in a built-in skill will not let you extract fields from documents in such languages. You can disable any of the supported languages, and we recommend disabling any languages that will not be used in your documents.
- For the Identity Document skill, Manual Review will not display the regions of the extracted fields.
Mobile Input
- The capture of passports is not currently automated. Users should manually capture the document image once the document has been centered on the screen.
Scanning Station
-
When installing the Scanning Station without administrative permissions, it will be installed with the English UI, even if a different UI language was selected during setup. To overcome this limitation, use one the following workarounds: (a) secure administrative permissions prior to installing the Scanning Station; (b) reinstall the Scanning Station; or (c) modify the following registry key after installing:
Where X equals:
- 64 for Chinese (Simplified)
- 3 for French
- 2 for German
- 6 for Italian
- 26 for Japanese
- 63 for Korean
- 27 for Portuguese (Brazil)
- 5 for Spanish