Known Issues and Workarounds
The following is a list of known issues discovered in this release for which a workaround is provided.
| Category | Known Issue | Workaround |
| AI Table Rule Builder | When using the Manual mode of the AI Table Rule Builder with the Microsoft Edge browser, the row separators do not render properly. | Resize the manual mode pane or click somewhere else on the page. This will temporarily allow you to see the final result. |
| AI Table Rule Builder | Standard Mode: If a user comes back from the Save Extraction Rule screen to the Ignore Rows screen, extraction data is not refreshed and the ignored rows are no longer visible. | While at the Ignore Row screen, click on the Previous button and then click Next. This will reload the results in the Ignore Row screen. |
| AI Table Rule Builder | When a table rule is created using Advanced mode without extraction data for a column, the ‘Column Header’ and ‘Column Data’ lassos shrinks and becomes unusable for such columns when the table rule is edited in Transact | While editing the rules in transact. Click the Reset Column button to reset the lassos for the column. |
| Installation | Installation may fail if VC++ 2015-2019 Redistributable is installed already. | To resolve this issue and proceed with the Transact installation, uninstall 2015-2019 VC++ redistributable. Once Transact is installed successfully, you may uninstall 2015 VC++ redistributable and restore the 2015-2019 VC++ redistributable. |
| Installation | When uninstalling Transact after upgrading to 2020.1.05, the version signature folder remains. | Manually delete the [EPHESOFT-HOME]\Version folder. |
| RecoStar for Linux | On importing a Windows batch class into Linux, RECOSTAR_HOCR language options won’t restore properly. | Select an alternate OCR Project file(rsp). Select original OCR project file(rsp). Set the proper language. Click Apply. |
| Reporting | The average page process graph under dashboard reports sometimes does not display, and instead presents the following error: “There was a problem creating this chart. Click here to view the details.” | This issue is isolated to some browsers. Use a different supported browser (Chrome, Edge) to view the average page process graph. |
| Upgrades | When updating from 2020.1 to 2020.1.04, the registration URL does not update which results in cloud-plugin.properties URL still pointing to the old non-multicountry registration link. | To resolve this issue, install the CHE 2020.1.1 service pack on the 2020.1 version of Transact and then upgrade to 2020.1.04. |
Known Issues
The following is a list of known issues discovered in this release for which there is not yet a workaround or fix.
| Category | Known Issue |
| AI Table Rule Builder | For Automated Mode: When two column headers are in a single column, unexpected extraction results may occur. |
| AI Table Rule Builder | For Automated Mode: If the spacing between two columns of data is insufficient, the data may merge into one column. |
| AI Table Rule Builder | Standard Mode: Preceding Text
When defining the lasso’d area for the Preceding Text configuration screen, only the first row of data within the lasso’d area is considered for Start Pattern RegEx on the Save Extraction Rule configuration screen. |
| AI Table Rule Builder | Standard Mode: Ignore Rows
For the Ignore Rows configuration screen, each selection that is composed of multiple rows of information (new row anchor data) may only remove the first of that grouped information. Note: This only happens if any column is marked with this selection: “This column spans a single row, and is always present.” |
| AI Table Rule Builder | Auto Table Extraction and detection will not extract data when switching between multiple modes. |
| AI Table Rule Builder | Swagger documentation is incomplete for the recently added auto-table-web-service-api: Ephesoft Auto Table Web Services. |
| AI Table Rule Builder | AI Table Rule Builder UI may freeze when processing a document with approximately 20 or more pages. |
| Cloud HyperExtender | When extracting handwritten text, the extraction accuracy may be low. You may experience the following problems:
1. The text around the ‘text of interest’ can get extracted along with the actual text. 2. The order of the extracted data may not be accurate for a multi-line extracted text. This is due to inaccurate HOCR coordinate information from Google Cloud Vision. our OCR engine provider. |
| Cloud HyperExtender | EText overlays in exported searchable PDFs are misaligned due to HOCR coordinate precision issues. |
| Cloud HyperExtender | In the tenant management dashboard, the Add User dialog box will close if the user clicks outside the dialog box. |
| Cloud HyperExtender | When using CHE to process a batch, batch instances intermittently fail and display the following error: “Could not execute cloud plugin.” |
| Cloud HyperExtender | When the CHE license is fully consumed (Remaining Pages is zero), HOCR operations dependent on CHE fail and return the error message “Internal server error.” This may occur while configuring extraction rules. |
| Cloud HyperExtender | Cloud HyperExtender will not return OCR results during batch class configuration when CHE Bring Your Own Key (BYOK) is utilized. The error message “Internal server error” displays.
Note: This error does not stop the batch instance workflow from executing successfully. |
| OCR Confidence | OCR confidence scores in the HOCR.xml file in Linux is lower than in Windows. Different Recostar versions are used between Windows and Linux and results will vary. |
| RecoStar for Linux | ADVANCED_BARCODE_EXTRACTION plugin is not compatible with the RECOSTAR_HOCR plugin. |
| RecoStar for Linux | Unexpected default values for RECOSTAR_HOCR plugin may show on first use in a batch class. Setting plugin values manually may be necessary. Default values can not be set. |
| RecoStar for Linux | When using the RECOSTAR_HOCR plugin on Linux, Chinese, Japanese, Korean, Thai, and Vietnamese cannot be combined with other languages. OCR operations may fail during configuration and at batch instance runtime. |
| RecoStar for Linux | EText is not supported for the RECOSTAR_HOCR plugin on Linux.
When the EText option is selected in the plugin, clicking the View OCR Data button results in the following error message: “Unable to perform OCRing of given file.” |
| Recostar for Linux | Due to coordinate issues, data may be extracted incorrectly when the RECOSTAR_HOCR plugin rotates or deskews ingested images. |